// Provides the "linkage" between a parse_node_tree_t and actual execution structures (job_t, etc.) // // A note on error handling: fish has two kind of errors, fatal parse errors non-fatal runtime // errors. A fatal error prevents execution of the entire file, while a non-fatal error skips that // job. // // Non-fatal errors are printed as soon as they are encountered; otherwise you would have to wait // for the execution to finish to see them. #include "config.h" // IWYU pragma: keep #include #include #include #include #include #include #include #include #include #include #include #include #include #include "builtin.h" #include "common.h" #include "complete.h" #include "env.h" #include "event.h" #include "exec.h" #include "expand.h" #include "function.h" #include "io.h" #include "parse_constants.h" #include "parse_execution.h" #include "parse_tree.h" #include "parse_util.h" #include "parser.h" #include "path.h" #include "proc.h" #include "reader.h" #include "tokenizer.h" #include "util.h" #include "wildcard.h" #include "wutil.h" /// These are the specific statement types that support redirections. static bool specific_statement_type_is_redirectable_block(const parse_node_t &node) { return node.type == symbol_block_statement || node.type == symbol_if_statement || node.type == symbol_switch_statement; } /// Get the name of a redirectable block, for profiling purposes. static wcstring profiling_cmd_name_for_redirectable_block(const parse_node_t &node, const parse_node_tree_t &tree, const wcstring &src) { assert(specific_statement_type_is_redirectable_block(node)); assert(node.has_source()); // Get the source for the block, and cut it at the next statement terminator. const size_t src_start = node.source_start; size_t src_len = node.source_length; const parse_node_tree_t::parse_node_list_t statement_terminator_nodes = tree.find_nodes(node, parse_token_type_end, 1); if (!statement_terminator_nodes.empty()) { const parse_node_t *term = statement_terminator_nodes.at(0); assert(term->source_start >= src_start); src_len = term->source_start - src_start; } wcstring result = wcstring(src, src_start, src_len); result.append(L"..."); return result; } parse_execution_context_t::parse_execution_context_t(moved_ref t, const wcstring &s, parser_t *p, int initial_eval_level) : tree(t), src(s), parser(p), eval_level(initial_eval_level), executing_node_idx(NODE_OFFSET_INVALID), cached_lineno_offset(0), cached_lineno_count(0) {} // Utilities wcstring parse_execution_context_t::get_source(const parse_node_t &node) const { return node.get_source(this->src); } const parse_node_t *parse_execution_context_t::get_child(const parse_node_t &parent, node_offset_t which, parse_token_type_t expected_type) const { return this->tree.get_child(parent, which, expected_type); } node_offset_t parse_execution_context_t::get_offset(const parse_node_t &node) const { // Get the offset of a node via pointer arithmetic, very hackish. const parse_node_t *addr = &node; const parse_node_t *base = &this->tree.at(0); assert(addr >= base); assert(addr - base < SOURCE_OFFSET_INVALID); node_offset_t offset = static_cast(addr - base); assert(offset < this->tree.size()); assert(&tree.at(offset) == &node); return offset; } const parse_node_t *parse_execution_context_t::infinite_recursive_statement_in_job_list( const parse_node_t &job_list, wcstring *out_func_name) const { assert(job_list.type == symbol_job_list); // This is a bit fragile. It is a test to see if we are inside of function call, but not inside // a block in that function call. If, in the future, the rules for what block scopes are pushed // on function invocation changes, then this check will break. const block_t *current = parser->block_at_index(0), *parent = parser->block_at_index(1); bool is_within_function_call = (current && parent && current->type() == TOP && parent->type() == FUNCTION_CALL); if (!is_within_function_call) { return NULL; } // Check to see which function call is forbidden. if (parser->forbidden_function.empty()) { return NULL; } const wcstring &forbidden_function_name = parser->forbidden_function.back(); // Get the first job in the job list. const parse_node_t *first_job = tree.next_node_in_node_list(job_list, symbol_job, NULL); if (first_job == NULL) { return NULL; } // Here's the statement node we find that's infinite recursive. const parse_node_t *infinite_recursive_statement = NULL; // Get the list of statements. const parse_node_tree_t::parse_node_list_t statements = tree.specific_statements_for_job(*first_job); // Find all the decorated statements. We are interested in statements with no decoration (i.e. // not command, not builtin) whose command expands to the forbidden function. for (size_t i = 0; i < statements.size(); i++) { // We only care about decorated statements, not while statements, etc. const parse_node_t &statement = *statements.at(i); if (statement.type != symbol_decorated_statement) { continue; } const parse_node_t &plain_statement = tree.find_child(statement, symbol_plain_statement); if (tree.decoration_for_plain_statement(plain_statement) != parse_statement_decoration_none) { // This statement has a decoration like 'builtin' or 'command', and therefore is not // infinite recursion. In particular this is what enables 'wrapper functions'. continue; } // Ok, this is an undecorated plain statement. Get and expand its command. wcstring cmd; tree.command_for_plain_statement(plain_statement, src, &cmd); if (expand_one(cmd, EXPAND_SKIP_CMDSUBST | EXPAND_SKIP_VARIABLES, NULL) && cmd == forbidden_function_name) { // This is it. infinite_recursive_statement = &statement; if (out_func_name != NULL) { *out_func_name = forbidden_function_name; } break; } } assert(infinite_recursive_statement == NULL || infinite_recursive_statement->type == symbol_decorated_statement); return infinite_recursive_statement; } enum process_type_t parse_execution_context_t::process_type_for_command( const parse_node_t &plain_statement, const wcstring &cmd) const { assert(plain_statement.type == symbol_plain_statement); enum process_type_t process_type = EXTERNAL; // Determine the process type, which depends on the statement decoration (command, builtin, // etc). enum parse_statement_decoration_t decoration = tree.decoration_for_plain_statement(plain_statement); if (decoration == parse_statement_decoration_exec) { // Always exec. process_type = INTERNAL_EXEC; } else if (decoration == parse_statement_decoration_command) { // Always a command. process_type = EXTERNAL; } else if (decoration == parse_statement_decoration_builtin) { // What happens if this builtin is not valid? process_type = INTERNAL_BUILTIN; } else if (function_exists(cmd)) { process_type = INTERNAL_FUNCTION; } else if (builtin_exists(cmd)) { process_type = INTERNAL_BUILTIN; } else { process_type = EXTERNAL; } return process_type; } bool parse_execution_context_t::should_cancel_execution(const block_t *block) const { return cancellation_reason(block) != execution_cancellation_none; } parse_execution_context_t::execution_cancellation_reason_t parse_execution_context_t::cancellation_reason(const block_t *block) const { if (shell_is_exiting()) { return execution_cancellation_exit; } if (parser && parser->cancellation_requested) { return execution_cancellation_skip; } if (block && block->loop_status != LOOP_NORMAL) { // Nasty hack - break and continue set the 'skip' flag as well as the loop status flag. return execution_cancellation_loop_control; } if (block && block->skip) { return execution_cancellation_skip; } return execution_cancellation_none; } /// Return whether the job contains a single statement, of block type, with no redirections. bool parse_execution_context_t::job_is_simple_block(const parse_node_t &job_node) const { assert(job_node.type == symbol_job); // Must have one statement. const parse_node_t &statement = *get_child(job_node, 0, symbol_statement); const parse_node_t &specific_statement = *get_child(statement, 0); if (!specific_statement_type_is_redirectable_block(specific_statement)) { // Not an appropriate block type. return false; } // Must be no pipes. const parse_node_t &continuation = *get_child(job_node, 1, symbol_job_continuation); if (continuation.child_count > 0) { // Multiple statements in this job, so there's pipes involved. return false; } // Check for arguments and redirections. All of the above types have an arguments / redirections // list. It must be empty. const parse_node_t &args_and_redirections = tree.find_child(specific_statement, symbol_arguments_or_redirections_list); if (args_and_redirections.child_count > 0) { // Non-empty, we have an argument or redirection. return false; } // Ok, we are a simple block! return true; } parse_execution_result_t parse_execution_context_t::run_if_statement( const parse_node_t &statement) { assert(statement.type == symbol_if_statement); // Push an if block. if_block_t *ib = new if_block_t(); ib->node_offset = this->get_offset(statement); parser->push_block(ib); parse_execution_result_t result = parse_execution_success; // We have a sequence of if clauses, with a final else, resulting in a single job list that we // execute. const parse_node_t *job_list_to_execute = NULL; const parse_node_t *if_clause = get_child(statement, 0, symbol_if_clause); const parse_node_t *else_clause = get_child(statement, 1, symbol_else_clause); for (;;) { if (should_cancel_execution(ib)) { result = parse_execution_cancelled; break; } // An if condition has a job and a "tail" of andor jobs, e.g. "foo ; and bar; or baz". assert(if_clause != NULL && else_clause != NULL); const parse_node_t &condition_head = *get_child(*if_clause, 1, symbol_job); const parse_node_t &condition_boolean_tail = *get_child(*if_clause, 3, symbol_andor_job_list); // Check the condition and the tail. We treat parse_execution_errored here as failure, in // accordance with historic behavior. parse_execution_result_t cond_ret = run_1_job(condition_head, ib); if (cond_ret == parse_execution_success) { cond_ret = run_job_list(condition_boolean_tail, ib); } const bool take_branch = (cond_ret == parse_execution_success) && proc_get_last_status() == EXIT_SUCCESS; if (take_branch) { // Condition succeeded. job_list_to_execute = get_child(*if_clause, 4, symbol_job_list); break; } else if (else_clause->child_count == 0) { // 'if' condition failed, no else clause, return 0, we're done. job_list_to_execute = NULL; proc_set_last_status(STATUS_BUILTIN_OK); break; } else { // We have an 'else continuation' (either else-if or else). const parse_node_t &else_cont = *get_child(*else_clause, 1, symbol_else_continuation); const parse_node_t *maybe_if_clause = get_child(else_cont, 0); if (maybe_if_clause && maybe_if_clause->type == symbol_if_clause) { // it's an 'else if', go to the next one. if_clause = maybe_if_clause; else_clause = get_child(else_cont, 1, symbol_else_clause); } else { // It's the final 'else', we're done. job_list_to_execute = get_child(else_cont, 1, symbol_job_list); break; } } } // Execute any job list we got. if (job_list_to_execute != NULL) { run_job_list(*job_list_to_execute, ib); } else { // No job list means no sucessful conditions, so return 0 (issue #1443). proc_set_last_status(STATUS_BUILTIN_OK); } // It's possible there's a last-minute cancellation (issue #1297). if (should_cancel_execution(ib)) { result = parse_execution_cancelled; } // Done parser->pop_block(ib); // Otherwise, take the exit status of the job list. Reversal of issue #1061. return result; } parse_execution_result_t parse_execution_context_t::run_begin_statement( const parse_node_t &header, const parse_node_t &contents) { assert(header.type == symbol_begin_header); assert(contents.type == symbol_job_list); // Basic begin/end block. Push a scope block. scope_block_t *sb = new scope_block_t(BEGIN); parser->push_block(sb); // Run the job list. parse_execution_result_t ret = run_job_list(contents, sb); // Pop the block. parser->pop_block(sb); return ret; } // Define a function. parse_execution_result_t parse_execution_context_t::run_function_statement( const parse_node_t &header, const parse_node_t &block_end_command) { assert(header.type == symbol_function_header); assert(block_end_command.type == symbol_end_command); // Get arguments. wcstring_list_t argument_list; parse_execution_result_t result = this->determine_arguments(header, &argument_list, failglob); if (result == parse_execution_success) { // The function definition extends from the end of the header to the function end. It's not // just the range of the contents because that loses comments - see issue #1710. assert(block_end_command.has_source()); size_t contents_start = header.source_start + header.source_length; size_t contents_end = block_end_command.source_start; // 1 past the last character in the function definition assert(contents_end >= contents_start); // Swallow whitespace at both ends. while (contents_start < contents_end && iswspace(this->src.at(contents_start))) { contents_start++; } while (contents_start < contents_end && iswspace(this->src.at(contents_end - 1))) { contents_end--; } assert(contents_end >= contents_start); const wcstring contents_str = wcstring(this->src, contents_start, contents_end - contents_start); int definition_line_offset = this->line_offset_of_character_at_offset(contents_start); wcstring error_str; io_streams_t streams; int err = builtin_function(*parser, streams, argument_list, contents_str, definition_line_offset, &error_str); proc_set_last_status(err); if (!error_str.empty()) { this->report_error(header, L"%ls", error_str.c_str()); result = parse_execution_errored; } } return result; } parse_execution_result_t parse_execution_context_t::run_block_statement( const parse_node_t &statement) { assert(statement.type == symbol_block_statement); const parse_node_t &block_header = *get_child(statement, 0, symbol_block_header); // block header const parse_node_t &header = *get_child(block_header, 0); // specific header type (e.g. for loop) const parse_node_t &contents = *get_child(statement, 1, symbol_job_list); // block contents parse_execution_result_t ret = parse_execution_success; switch (header.type) { case symbol_for_header: { ret = run_for_statement(header, contents); break; } case symbol_while_header: { ret = run_while_statement(header, contents); break; } case symbol_function_header: { const parse_node_t &function_end = *get_child( statement, 2, symbol_end_command); // the 'end' associated with the block ret = run_function_statement(header, function_end); break; } case symbol_begin_header: { ret = run_begin_statement(header, contents); break; } default: { fprintf(stderr, "Unexpected block header: %ls\n", header.describe().c_str()); PARSER_DIE(); break; } } return ret; } parse_execution_result_t parse_execution_context_t::run_for_statement( const parse_node_t &header, const parse_node_t &block_contents) { assert(header.type == symbol_for_header); assert(block_contents.type == symbol_job_list); // Get the variable name: `for var_name in ...`. We expand the variable name. It better result // in just one. const parse_node_t &var_name_node = *get_child(header, 1, parse_token_type_string); wcstring for_var_name = get_source(var_name_node); if (!expand_one(for_var_name, 0, NULL)) { report_error(var_name_node, FAILED_EXPANSION_VARIABLE_NAME_ERR_MSG, for_var_name.c_str()); return parse_execution_errored; } // Get the contents to iterate over. wcstring_list_t argument_sequence; parse_execution_result_t ret = this->determine_arguments(header, &argument_sequence, nullglob); if (ret != parse_execution_success) { return ret; } for_block_t *fb = new for_block_t(); parser->push_block(fb); // Now drive the for loop. const size_t arg_count = argument_sequence.size(); for (size_t i = 0; i < arg_count; i++) { if (should_cancel_execution(fb)) { ret = parse_execution_cancelled; break; } const wcstring &val = argument_sequence.at(i); env_set(for_var_name, val.c_str(), ENV_LOCAL); fb->loop_status = LOOP_NORMAL; fb->skip = 0; this->run_job_list(block_contents, fb); if (this->cancellation_reason(fb) == execution_cancellation_loop_control) { // Handle break or continue. if (fb->loop_status == LOOP_CONTINUE) { // Reset the loop state. fb->loop_status = LOOP_NORMAL; fb->skip = false; continue; } else if (fb->loop_status == LOOP_BREAK) { break; } } } parser->pop_block(fb); return ret; } parse_execution_result_t parse_execution_context_t::run_switch_statement( const parse_node_t &statement) { assert(statement.type == symbol_switch_statement); parse_execution_result_t result = parse_execution_success; // Get the switch variable. const parse_node_t &switch_value_node = *get_child(statement, 1, symbol_argument); const wcstring switch_value = get_source(switch_value_node); // Expand it. We need to offset any errors by the position of the string. std::vector switch_values_expanded; parse_error_list_t errors; int expand_ret = expand_string(switch_value, &switch_values_expanded, EXPAND_NO_DESCRIPTIONS, &errors); parse_error_offset_source_start(&errors, switch_value_node.source_start); switch (expand_ret) { case EXPAND_ERROR: { result = report_errors(errors); break; } case EXPAND_WILDCARD_NO_MATCH: { result = report_unmatched_wildcard_error(switch_value_node); break; } case EXPAND_WILDCARD_MATCH: case EXPAND_OK: { break; } } if (result == parse_execution_success && switch_values_expanded.size() != 1) { result = report_error(switch_value_node, _(L"switch: Expected exactly one argument, got %lu\n"), switch_values_expanded.size()); } if (result == parse_execution_success) { const wcstring &switch_value_expanded = switch_values_expanded.at(0).completion; switch_block_t *sb = new switch_block_t(); parser->push_block(sb); // Expand case statements. const parse_node_t *case_item_list = get_child(statement, 3, symbol_case_item_list); // Loop while we don't have a match but do have more of the list. const parse_node_t *matching_case_item = NULL; while (matching_case_item == NULL && case_item_list != NULL) { if (should_cancel_execution(sb)) { result = parse_execution_cancelled; break; } // Get the next item and the remainder of the list. const parse_node_t *case_item = tree.next_node_in_node_list(*case_item_list, symbol_case_item, &case_item_list); if (case_item == NULL) { // No more items. break; } // Pull out the argument list. const parse_node_t &arg_list = *get_child(*case_item, 1, symbol_argument_list); // Expand arguments. A case item list may have a wildcard that fails to expand to // anything. We also report case errors, but don't stop execution; i.e. a case item that // contains an unexpandable process will report and then fail to match. wcstring_list_t case_args; parse_execution_result_t case_result = this->determine_arguments(arg_list, &case_args, failglob); if (case_result == parse_execution_success) { for (size_t i = 0; i < case_args.size(); i++) { const wcstring &arg = case_args.at(i); // Unescape wildcards so they can be expanded again. wcstring unescaped_arg = parse_util_unescape_wildcards(arg); bool match = wildcard_match(switch_value_expanded, unescaped_arg); // If this matched, we're done. if (match) { matching_case_item = case_item; break; } } } } if (result == parse_execution_success && matching_case_item != NULL) { // Success, evaluate the job list. const parse_node_t *job_list = get_child(*matching_case_item, 3, symbol_job_list); result = this->run_job_list(*job_list, sb); } parser->pop_block(sb); } return result; } parse_execution_result_t parse_execution_context_t::run_while_statement( const parse_node_t &header, const parse_node_t &block_contents) { assert(header.type == symbol_while_header); assert(block_contents.type == symbol_job_list); // Push a while block. while_block_t *wb = new while_block_t(); wb->node_offset = this->get_offset(header); parser->push_block(wb); parse_execution_result_t ret = parse_execution_success; // The conditions of the while loop. const parse_node_t &condition_head = *get_child(header, 1, symbol_job); const parse_node_t &condition_boolean_tail = *get_child(header, 3, symbol_andor_job_list); // Run while the condition is true. for (;;) { // Check the condition. parse_execution_result_t cond_ret = this->run_1_job(condition_head, wb); if (cond_ret == parse_execution_success) { cond_ret = run_job_list(condition_boolean_tail, wb); } // We only continue on successful execution and EXIT_SUCCESS. if (cond_ret != parse_execution_success || proc_get_last_status() != EXIT_SUCCESS) { break; } // Check cancellation. if (this->should_cancel_execution(wb)) { ret = parse_execution_cancelled; break; } // The block ought to go inside the loop (see issue #1212). this->run_job_list(block_contents, wb); if (this->cancellation_reason(wb) == execution_cancellation_loop_control) { // Handle break or continue. if (wb->loop_status == LOOP_CONTINUE) { // Reset the loop state. wb->loop_status = LOOP_NORMAL; wb->skip = false; continue; } else if (wb->loop_status == LOOP_BREAK) { break; } } // no_exec means that fish was invoked with -n or --no-execute. If set, we allow the loop to // not-execute once so its contents can be checked, and then break. if (no_exec) { break; } } /* Done */ parser->pop_block(wb); return ret; } // Reports an error. Always returns parse_execution_errored, so you can assign the result to an // 'errored' variable. parse_execution_result_t parse_execution_context_t::report_error(const parse_node_t &node, const wchar_t *fmt, ...) const { // Create an error. parse_error_list_t error_list = parse_error_list_t(1); parse_error_t *error = &error_list.at(0); error->source_start = node.source_start; error->source_length = node.source_length; error->code = parse_error_syntax; // hackish va_list va; va_start(va, fmt); error->text = vformat_string(fmt, va); va_end(va); this->report_errors(error_list); return parse_execution_errored; } parse_execution_result_t parse_execution_context_t::report_errors( const parse_error_list_t &error_list) const { if (!parser->cancellation_requested) { if (error_list.empty()) { fprintf(stderr, "Bug: Error reported but no error text found."); } // Get a backtrace. wcstring backtrace_and_desc; parser->get_backtrace(src, error_list, &backtrace_and_desc); // Print it. fprintf(stderr, "%ls", backtrace_and_desc.c_str()); } return parse_execution_errored; } /// Reports an unmatched wildcard error and returns parse_execution_errored. parse_execution_result_t parse_execution_context_t::report_unmatched_wildcard_error( const parse_node_t &unmatched_wildcard) { proc_set_last_status(STATUS_UNMATCHED_WILDCARD); report_error(unmatched_wildcard, WILDCARD_ERR_MSG, get_source(unmatched_wildcard).c_str()); return parse_execution_errored; } // Given a command string that might contain fish special tokens return a string without those // tokens. static wcstring reconstruct_orig_cmd(wcstring cmd_str) { // TODO(krader1961): Figure out what VARIABLE_EXPAND means in this context. After looking at the // code and doing various tests I couldn't figure out why that token would be present when this // code is run. I was therefore unable to determine how to substitute its presence in the error // message. wcstring orig_cmd = cmd_str; if (cmd_str.find(VARIABLE_EXPAND_SINGLE) != std::string::npos) { // Variable was quoted to force expansion of multiple elements into a single element. // // The following isn't entirely correct. For example, $abc"$def" will become "$abc$def". // However, anyone writing the former is asking for trouble so I don't feel bad about not // accurately reconstructing what they typed. wcstring new_cmd_str = wcstring(cmd_str); std::replace(new_cmd_str.begin(), new_cmd_str.end(), (wchar_t)VARIABLE_EXPAND_SINGLE, L'$'); orig_cmd = L"\"" + new_cmd_str + L"\""; } return orig_cmd; } /// Handle the case of command not found. parse_execution_result_t parse_execution_context_t::handle_command_not_found( const wcstring &cmd_str, const parse_node_t &statement_node, int err_code) { assert(statement_node.type == symbol_plain_statement); // We couldn't find the specified command. This is a non-fatal error. We want to set the exit // status to 127, which is the standard number used by other shells like bash and zsh. const wchar_t *const cmd = cmd_str.c_str(); const wchar_t *const equals_ptr = wcschr(cmd, L'='); if (equals_ptr != NULL) { // Try to figure out if this is a pure variable assignment (foo=bar), or if this appears to // be running a command (foo=bar ruby...). const wcstring name_str = wcstring(cmd, equals_ptr - cmd); // variable name, up to the = const wcstring val_str = wcstring(equals_ptr + 1); // variable value, past the = const parse_node_tree_t::parse_node_list_t args = tree.find_nodes(statement_node, symbol_argument, 1); if (!args.empty()) { const wcstring argument = get_source(*args.at(0)); wcstring ellipsis_str = wcstring(1, ellipsis_char); if (ellipsis_str == L"$") ellipsis_str = L"..."; // Looks like a command. this->report_error(statement_node, ERROR_BAD_EQUALS_IN_COMMAND5, argument.c_str(), name_str.c_str(), val_str.c_str(), argument.c_str(), ellipsis_str.c_str()); } else { this->report_error(statement_node, ERROR_BAD_COMMAND_ASSIGN_ERR_MSG, name_str.c_str(), val_str.c_str()); } } else if (wcschr(cmd, L'$') || wcschr(cmd, VARIABLE_EXPAND_SINGLE) || wcschr(cmd, VARIABLE_EXPAND)) { wcstring eval_cmd = reconstruct_orig_cmd(cmd_str); this->report_error(statement_node, _(L"Variables may not be used as commands. In fish, " L"please define a function or use 'eval %ls'."), eval_cmd.c_str()); } else if (err_code != ENOENT) { this->report_error(statement_node, _(L"The file '%ls' is not executable by this user"), cmd ? cmd : L"UNKNOWN"); } else { // Handle unrecognized commands with standard command not found handler that can make better // error messages. wcstring_list_t event_args; { parse_execution_result_t arg_result = this->determine_arguments(statement_node, &event_args, failglob); if (arg_result != parse_execution_success) { return arg_result; } event_args.insert(event_args.begin(), cmd_str); } event_fire_generic(L"fish_command_not_found", &event_args); // Here we want to report an error (so it shows a backtrace), but with no text. this->report_error(statement_node, L""); } // Set the last proc status appropriately. proc_set_last_status(err_code == ENOENT ? STATUS_UNKNOWN_COMMAND : STATUS_NOT_EXECUTABLE); return parse_execution_errored; } /// Creates a 'normal' (non-block) process. parse_execution_result_t parse_execution_context_t::populate_plain_process( job_t *job, process_t *proc, const parse_node_t &statement) { assert(job != NULL); assert(proc != NULL); assert(statement.type == symbol_plain_statement); // We may decide that a command should be an implicit cd. bool use_implicit_cd = false; // Get the command. We expect to always get it here. wcstring cmd; bool got_cmd = tree.command_for_plain_statement(statement, src, &cmd); assert(got_cmd); // Expand it as a command. Return an error on failure. bool expanded = expand_one(cmd, EXPAND_SKIP_CMDSUBST | EXPAND_SKIP_VARIABLES, NULL); if (!expanded) { report_error(statement, ILLEGAL_CMD_ERR_MSG, cmd.c_str()); return parse_execution_errored; } // Determine the process type. enum process_type_t process_type = process_type_for_command(statement, cmd); // Check for stack overflow. if (process_type == INTERNAL_FUNCTION && parser->forbidden_function.size() > FISH_MAX_STACK_DEPTH) { this->report_error(statement, CALL_STACK_LIMIT_EXCEEDED_ERR_MSG); return parse_execution_errored; } wcstring path_to_external_command; if (process_type == EXTERNAL || process_type == INTERNAL_EXEC) { // Determine the actual command. This may be an implicit cd. bool has_command = path_get_path(cmd, &path_to_external_command); // If there was no command, then we care about the value of errno after checking for it, to // distinguish between e.g. no file vs permissions problem. const int no_cmd_err_code = errno; // If the specified command does not exist, and is undecorated, try using an implicit cd. if (!has_command && tree.decoration_for_plain_statement(statement) == parse_statement_decoration_none) { // Implicit cd requires an empty argument and redirection list. const parse_node_t *args = get_child(statement, 1, symbol_arguments_or_redirections_list); if (args->child_count == 0) { // Ok, no arguments or redirections; check to see if the first argument is a // directory. wcstring implicit_cd_path; use_implicit_cd = path_can_be_implicit_cd(cmd, &implicit_cd_path); } } if (!has_command && !use_implicit_cd) { // No command. return this->handle_command_not_found(cmd, statement, no_cmd_err_code); } } // The argument list and set of IO redirections that we will construct for the process. io_chain_t process_io_chain; wcstring_list_t argument_list; if (use_implicit_cd) { /* Implicit cd is simple */ argument_list.push_back(L"cd"); argument_list.push_back(cmd); path_to_external_command.clear(); // If we have defined a wrapper around cd, use it, otherwise use the cd builtin. process_type = function_exists(L"cd") ? INTERNAL_FUNCTION : INTERNAL_BUILTIN; } else { const globspec_t glob_behavior = contains(cmd, L"set", L"count") ? nullglob : failglob; // Form the list of arguments. The command is the first argument. TODO: count hack, where we // treat 'count --help' as different from 'count $foo' that expands to 'count --help'. fish // 1.x never successfully did this, but it tried to! parse_execution_result_t arg_result = this->determine_arguments(statement, &argument_list, glob_behavior); if (arg_result != parse_execution_success) { return arg_result; } argument_list.insert(argument_list.begin(), cmd); // The set of IO redirections that we construct for the process. if (!this->determine_io_chain(statement, &process_io_chain)) { return parse_execution_errored; } // Determine the process type. process_type = process_type_for_command(statement, cmd); } // Populate the process. proc->type = process_type; proc->set_argv(argument_list); proc->set_io_chain(process_io_chain); proc->actual_cmd = path_to_external_command; return parse_execution_success; } // Determine the list of arguments, expanding stuff. Reports any errors caused by expansion. If we // have a wildcard that could not be expanded, report the error and continue. parse_execution_result_t parse_execution_context_t::determine_arguments( const parse_node_t &parent, wcstring_list_t *out_arguments, globspec_t glob_behavior) { // Get all argument nodes underneath the statement. We guess we'll have that many arguments (but // may have more or fewer, if there are wildcards involved). const parse_node_tree_t::parse_node_list_t argument_nodes = tree.find_nodes(parent, symbol_argument); out_arguments->reserve(out_arguments->size() + argument_nodes.size()); std::vector arg_expanded; for (size_t i = 0; i < argument_nodes.size(); i++) { const parse_node_t &arg_node = *argument_nodes.at(i); // Expect all arguments to have source. assert(arg_node.has_source()); const wcstring arg_str = arg_node.get_source(src); // Expand this string. parse_error_list_t errors; arg_expanded.clear(); int expand_ret = expand_string(arg_str, &arg_expanded, EXPAND_NO_DESCRIPTIONS, &errors); parse_error_offset_source_start(&errors, arg_node.source_start); switch (expand_ret) { case EXPAND_ERROR: { this->report_errors(errors); return parse_execution_errored; } case EXPAND_WILDCARD_NO_MATCH: { if (glob_behavior == failglob) { // Report the unmatched wildcard error and stop processing. report_unmatched_wildcard_error(arg_node); return parse_execution_errored; } break; } case EXPAND_WILDCARD_MATCH: case EXPAND_OK: { break; } } // Now copy over any expanded arguments. Do it using swap() to avoid extra allocations; this // is called very frequently. size_t old_arg_count = out_arguments->size(); size_t new_arg_count = arg_expanded.size(); out_arguments->resize(old_arg_count + new_arg_count); for (size_t i = 0; i < new_arg_count; i++) { wcstring &new_arg = arg_expanded.at(i).completion; out_arguments->at(old_arg_count + i).swap(new_arg); } } return parse_execution_success; } bool parse_execution_context_t::determine_io_chain(const parse_node_t &statement_node, io_chain_t *out_chain) { io_chain_t result; bool errored = false; // We are called with a statement of varying types. We require that the statement have an // arguments_or_redirections_list child. const parse_node_t &args_and_redirections_list = tree.find_child(statement_node, symbol_arguments_or_redirections_list); // Get all redirection nodes underneath the statement. const parse_node_tree_t::parse_node_list_t redirect_nodes = tree.find_nodes(args_and_redirections_list, symbol_redirection); for (size_t i = 0; i < redirect_nodes.size(); i++) { const parse_node_t &redirect_node = *redirect_nodes.at(i); int source_fd = -1; // source fd wcstring target; // file path or target fd enum token_type redirect_type = tree.type_for_redirection(redirect_node, src, &source_fd, &target); // PCA: I can't justify this EXPAND_SKIP_VARIABLES flag. It was like this when I got here. bool target_expanded = expand_one(target, no_exec ? EXPAND_SKIP_VARIABLES : 0, NULL); if (!target_expanded || target.empty()) { // TODO: Improve this error message. errored = report_error(redirect_node, _(L"Invalid redirection target: %ls"), target.c_str()); } // Generate the actual IO redirection. shared_ptr new_io; assert(redirect_type != TOK_NONE); switch (redirect_type) { case TOK_REDIRECT_FD: { if (target == L"-") { new_io.reset(new io_close_t(source_fd)); } else { wchar_t *end = NULL; errno = 0; int old_fd = fish_wcstoi(target.c_str(), &end, 10); if (old_fd < 0 || errno || *end) { errored = report_error(redirect_node, _(L"Requested redirection to '%ls', which " L"is not a valid file descriptor"), target.c_str()); } else { new_io.reset(new io_fd_t(source_fd, old_fd, true)); } } break; } case TOK_REDIRECT_OUT: case TOK_REDIRECT_APPEND: case TOK_REDIRECT_IN: case TOK_REDIRECT_NOCLOB: { int oflags = oflags_for_redirection_type(redirect_type); io_file_t *new_io_file = new io_file_t(source_fd, target, oflags); new_io.reset(new_io_file); break; } default: { // Should be unreachable. fprintf(stderr, "Unexpected redirection type %ld. aborting.\n", (long)redirect_type); PARSER_DIE(); break; } } // Append the new_io if we got one. if (new_io.get() != NULL) { result.push_back(new_io); } } if (out_chain && !errored) { out_chain->swap(result); } return !errored; } parse_execution_result_t parse_execution_context_t::populate_boolean_process( job_t *job, process_t *proc, const parse_node_t &bool_statement) { // Handle a boolean statement. bool skip_job = false; assert(bool_statement.type == symbol_boolean_statement); switch (parse_node_tree_t::statement_boolean_type(bool_statement)) { case parse_bool_and: { // AND. Skip if the last job failed. skip_job = (proc_get_last_status() != 0); break; } case parse_bool_or: { // OR. Skip if the last job succeeded. skip_job = (proc_get_last_status() == 0); break; } case parse_bool_not: { // NOT. Negate it. job_set_flag(job, JOB_NEGATE, !job_get_flag(job, JOB_NEGATE)); break; } } if (skip_job) { return parse_execution_skipped; } const parse_node_t &subject = *tree.get_child(bool_statement, 1, symbol_statement); return this->populate_job_process(job, proc, subject); } parse_execution_result_t parse_execution_context_t::populate_block_process( job_t *job, process_t *proc, const parse_node_t &statement_node) { // We handle block statements by creating INTERNAL_BLOCK_NODE, that will bounce back to us when // it's time to execute them. assert(statement_node.type == symbol_block_statement || statement_node.type == symbol_if_statement || statement_node.type == symbol_switch_statement); // The set of IO redirections that we construct for the process. io_chain_t process_io_chain; bool errored = !this->determine_io_chain(statement_node, &process_io_chain); if (errored) return parse_execution_errored; proc->type = INTERNAL_BLOCK_NODE; proc->internal_block_node = this->get_offset(statement_node); proc->set_io_chain(process_io_chain); return parse_execution_success; } // Returns a process_t allocated with new. It's the caller's responsibility to delete it (!). parse_execution_result_t parse_execution_context_t::populate_job_process( job_t *job, process_t *proc, const parse_node_t &statement_node) { assert(statement_node.type == symbol_statement); assert(statement_node.child_count == 1); // Get the "specific statement" which is boolean / block / if / switch / decorated. const parse_node_t &specific_statement = *get_child(statement_node, 0); parse_execution_result_t result = parse_execution_success; switch (specific_statement.type) { case symbol_boolean_statement: { result = this->populate_boolean_process(job, proc, specific_statement); break; } case symbol_block_statement: case symbol_if_statement: case symbol_switch_statement: { result = this->populate_block_process(job, proc, specific_statement); break; } case symbol_decorated_statement: { // Get the plain statement. It will pull out the decoration itself. const parse_node_t &plain_statement = tree.find_child(specific_statement, symbol_plain_statement); result = this->populate_plain_process(job, proc, plain_statement); break; } default: { fprintf(stderr, "'%ls' not handled by new parser yet\n", specific_statement.describe().c_str()); PARSER_DIE(); break; } } return result; } parse_execution_result_t parse_execution_context_t::populate_job_from_job_node( job_t *j, const parse_node_t &job_node, const block_t *associated_block) { assert(job_node.type == symbol_job); // Tell the job what its command is. j->set_command(get_source(job_node)); // We are going to construct process_t structures for every statement in the job. Get the first // statement. const parse_node_t *statement_node = get_child(job_node, 0, symbol_statement); assert(statement_node != NULL); parse_execution_result_t result = parse_execution_success; // Create processes. Each one may fail. std::vector processes; processes.push_back(new process_t()); result = this->populate_job_process(j, processes.back(), *statement_node); // Construct process_ts for job continuations (pipelines), by walking the list until we hit the // terminal (empty) job continuation. const parse_node_t *job_cont = get_child(job_node, 1, symbol_job_continuation); assert(job_cont != NULL); while (result == parse_execution_success && job_cont->child_count > 0) { assert(job_cont->type == symbol_job_continuation); // Handle the pipe, whose fd may not be the obvious stdout. const parse_node_t &pipe_node = *get_child(*job_cont, 0, parse_token_type_pipe); int pipe_write_fd = fd_redirected_by_pipe(get_source(pipe_node)); if (pipe_write_fd == -1) { result = report_error(pipe_node, ILLEGAL_FD_ERR_MSG, get_source(pipe_node).c_str()); break; } processes.back()->pipe_write_fd = pipe_write_fd; // Get the statement node and make a process from it. const parse_node_t *statement_node = get_child(*job_cont, 1, symbol_statement); assert(statement_node != NULL); // Store the new process (and maybe with an error). processes.push_back(new process_t()); result = this->populate_job_process(j, processes.back(), *statement_node); // Get the next continuation. job_cont = get_child(*job_cont, 2, symbol_job_continuation); assert(job_cont != NULL); } // Return what happened. if (result == parse_execution_success) { // Link up the processes. assert(!processes.empty()); j->first_process = processes.at(0); for (size_t i = 1; i < processes.size(); i++) { processes.at(i - 1)->next = processes.at(i); } } else { // Clean up processes. for (size_t i = 0; i < processes.size(); i++) { const process_t *proc = processes.at(i); processes.at(i) = NULL; delete proc; } } return result; } parse_execution_result_t parse_execution_context_t::run_1_job(const parse_node_t &job_node, const block_t *associated_block) { if (should_cancel_execution(associated_block)) { return parse_execution_cancelled; } // Get terminal modes. struct termios tmodes = {}; if (shell_is_interactive()) { if (tcgetattr(STDIN_FILENO, &tmodes)) { // Need real error handling here. wperror(L"tcgetattr"); return parse_execution_errored; } } // Increment the eval_level for the duration of this command. scoped_push saved_eval_level(&eval_level, eval_level + 1); // Save the node index. scoped_push saved_node_offset(&executing_node_idx, this->get_offset(job_node)); // Profiling support. long long start_time = 0, parse_time = 0, exec_time = 0; profile_item_t *profile_item = this->parser->create_profile_item(); if (profile_item != NULL) { start_time = get_time(); } // When we encounter a block construct (e.g. while loop) in the general case, we create a "block // process" that has a pointer to its source. This allows us to handle block-level redirections. // However, if there are no redirections, then we can just jump into the block directly, which // is significantly faster. if (job_is_simple_block(job_node)) { parse_execution_result_t result = parse_execution_success; const parse_node_t &statement = *get_child(job_node, 0, symbol_statement); const parse_node_t &specific_statement = *get_child(statement, 0); assert(specific_statement_type_is_redirectable_block(specific_statement)); switch (specific_statement.type) { case symbol_block_statement: { result = this->run_block_statement(specific_statement); break; } case symbol_if_statement: { result = this->run_if_statement(specific_statement); break; } case symbol_switch_statement: { result = this->run_switch_statement(specific_statement); break; } default: { // Other types should be impossible due to the // specific_statement_type_is_redirectable_block check. PARSER_DIE(); break; } } if (profile_item != NULL) { // Block-types profile a little weird. They have no 'parse' time, and their command is // just the block type. exec_time = get_time(); profile_item->level = eval_level; profile_item->parse = 0; profile_item->exec = (int)(exec_time - start_time); profile_item->cmd = profiling_cmd_name_for_redirectable_block(specific_statement, this->tree, this->src); profile_item->skipped = false; } return result; } job_t *j = new job_t(acquire_job_id(), block_io); j->tmodes = tmodes; job_set_flag(j, JOB_CONTROL, (job_control_mode == JOB_CONTROL_ALL) || ((job_control_mode == JOB_CONTROL_INTERACTIVE) && (shell_is_interactive()))); job_set_flag(j, JOB_FOREGROUND, !tree.job_should_be_backgrounded(job_node)); job_set_flag(j, JOB_TERMINAL, job_get_flag(j, JOB_CONTROL) && !is_subshell && !is_event); job_set_flag(j, JOB_SKIP_NOTIFICATION, is_subshell || is_block || is_event || !shell_is_interactive()); // Tell the current block what its job is. This has to happen before we populate it (#1394). parser->current_block()->job = j; // Populate the job. This may fail for reasons like command_not_found. If this fails, an error // will have been printed. parse_execution_result_t pop_result = this->populate_job_from_job_node(j, job_node, associated_block); // Clean up the job on failure or cancellation. bool populated_job = (pop_result == parse_execution_success); if (!populated_job || this->should_cancel_execution(associated_block)) { assert(parser->current_block()->job == j); parser->current_block()->job = NULL; delete j; j = NULL; populated_job = false; } // Store time it took to 'parse' the command. if (profile_item != NULL) { parse_time = get_time(); } if (populated_job) { // Success. Give the job to the parser - it will clean it up. parser->job_add(j); // Check to see if this contained any external commands. bool job_contained_external_command = false; for (const process_t *proc = j->first_process; proc != NULL; proc = proc->next) { if (proc->type == EXTERNAL) { job_contained_external_command = true; break; } } // Actually execute the job. exec_job(*this->parser, j); // Only external commands require a new fishd barrier. if (job_contained_external_command) { set_proc_had_barrier(false); } } if (profile_item != NULL) { exec_time = get_time(); profile_item->level = eval_level; profile_item->parse = (int)(parse_time - start_time); profile_item->exec = (int)(exec_time - parse_time); profile_item->cmd = j ? j->command() : wcstring(); profile_item->skipped = !populated_job; } job_reap(0); // clean up jobs return parse_execution_success; } parse_execution_result_t parse_execution_context_t::run_job_list(const parse_node_t &job_list_node, const block_t *associated_block) { assert(job_list_node.type == symbol_job_list || job_list_node.type == symbol_andor_job_list); parse_execution_result_t result = parse_execution_success; const parse_node_t *job_list = &job_list_node; while (job_list != NULL && !should_cancel_execution(associated_block)) { assert(job_list->type == symbol_job_list || job_list_node.type == symbol_andor_job_list); // Try pulling out a job. const parse_node_t *job = tree.next_node_in_node_list(*job_list, symbol_job, &job_list); if (job != NULL) { result = this->run_1_job(*job, associated_block); } } // Returns the last job executed. return result; } parse_execution_result_t parse_execution_context_t::eval_node_at_offset( node_offset_t offset, const block_t *associated_block, const io_chain_t &io) { // Don't ever expect to have an empty tree if this is called. assert(!tree.empty()); assert(offset < tree.size()); // Apply this block IO for the duration of this function. scoped_push block_io_push(&block_io, io); const parse_node_t &node = tree.at(offset); // Currently, we only expect to execute the top level job list, or a block node. Assert that. assert(node.type == symbol_job_list || specific_statement_type_is_redirectable_block(node)); enum parse_execution_result_t status = parse_execution_success; switch (node.type) { case symbol_job_list: { // We should only get a job list if it's the very first node. This is because this is // the entry point for both top-level execution (the first node) and INTERNAL_BLOCK_NODE // execution (which does block statements, but never job lists). assert(offset == 0); wcstring func_name; const parse_node_t *infinite_recursive_node = this->infinite_recursive_statement_in_job_list(node, &func_name); if (infinite_recursive_node != NULL) { // We have an infinite recursion. this->report_error(*infinite_recursive_node, INFINITE_FUNC_RECURSION_ERR_MSG, func_name.c_str()); status = parse_execution_errored; } else { // No infinite recursion. status = this->run_job_list(node, associated_block); } break; } case symbol_block_statement: { status = this->run_block_statement(node); break; } case symbol_if_statement: { status = this->run_if_statement(node); break; } case symbol_switch_statement: { status = this->run_switch_statement(node); break; } default: { // In principle, we could support other node types. However we never expect to be passed // them - see above. fprintf(stderr, "Unexpected node %ls found in %s\n", node.describe().c_str(), __FUNCTION__); PARSER_DIE(); break; } } return status; } int parse_execution_context_t::line_offset_of_node_at_offset(node_offset_t requested_index) { // If we're not executing anything, return -1. if (requested_index == NODE_OFFSET_INVALID) { return -1; } // If for some reason we're executing a node without source, return -1. const parse_node_t &node = tree.at(requested_index); if (!node.has_source()) { return -1; } size_t char_offset = tree.at(requested_index).source_start; return this->line_offset_of_character_at_offset(char_offset); } int parse_execution_context_t::line_offset_of_character_at_offset(size_t offset) { // Count the number of newlines, leveraging our cache. assert(offset <= src.size()); // Easy hack to handle 0. if (offset == 0) { return 0; } // We want to return (one plus) the number of newlines at offsets less than the given offset. // cached_lineno_count is the number of newlines at indexes less than cached_lineno_offset. const wchar_t *str = src.c_str(); if (offset > cached_lineno_offset) { size_t i; for (i = cached_lineno_offset; str[i] != L'\0' && i < offset; i++) { // Add one for every newline we find in the range [cached_lineno_offset, offset). if (str[i] == L'\n') { cached_lineno_count++; } } cached_lineno_offset = i; // note: i, not offset, in case offset is beyond the length of the string } else if (offset < cached_lineno_offset) { // Subtract one for every newline we find in the range [offset, cached_lineno_offset). for (size_t i = offset; i < cached_lineno_offset; i++) { if (str[i] == L'\n') { cached_lineno_count--; } } cached_lineno_offset = offset; } return cached_lineno_count; } int parse_execution_context_t::get_current_line_number() { int line_number = -1; int line_offset = this->line_offset_of_node_at_offset(this->executing_node_idx); if (line_offset >= 0) { // The offset is 0 based; the number is 1 based. line_number = line_offset + 1; } return line_number; } int parse_execution_context_t::get_current_source_offset() const { int result = -1; if (executing_node_idx != NODE_OFFSET_INVALID) { const parse_node_t &node = tree.at(executing_node_idx); if (node.has_source()) { result = static_cast(node.source_start); } } return result; }