* Move handling of headers around to prevent memory use spikes
While debugging #782, I noticed that using http2 and max_fails=0,
X-Forwarded-For grew infinitely when an upstream request failed after
refreshing the test page. This change ensures that headers are only
set once per request rather than appending in a time-terminated loop.
* Refactor some code into its own function
[1] 57e459e02b/src/crypto/tls/common.go (L424)
[2] 57e459e02b/src/crypto/tls/common.go (L392-L407)
[2] has overwritten the first tls ticket key on round N=0, that has previously
been written using [1].
Go's stdlib does not use c.sessionTicketKeys≥1 as indicator if those values had
already been set; initializing that lone SessionTicketKey does the job for for
now.
If c.serverInit() were called in round N+1 all existing tls ticket keys
would be overwritten (in round N<4 except the very first one, of course).
As member variables of tls.Config are read-only by then, we cannot keep
updating SessionTicketKey as well.
This has been escalated to Go's authors with golang/go#15421 here:
https://github.com/golang/go/issues/15421
Thanks to Matthew Holt for the initial report!
Those settings enforce convergence on common coding style with respect to whitespace.
Do not use tabs to indent with shell scripts because those tabs most often
serve the function of triggering command completion. Which could end a
command before it is pasted completely.
Traditionally indentation is two spaces here, not four.
Other rules will catch stray whitespace at the end of lines or files, which,
once committed, would annoy the next developer because his editor would strip
them from lines he did not intended to modify in the first place.
Fixes a surplus — next to "go up".
Identifies the preamble as the table's summary.
Emits filesizes in bytes, which can be consumed by any browser-side scripts
or utilized in sorting when the table is copy-and-pasted into a spreadsheet
software.
Uses <time> along with proper datetime representation, which a browser could
utilize to display the datetime rendered according to the requestor's locale.
Assigns negative sizes to directories in order to have them listed reliably
before any zero-sized files. That order is what most users expect when
sorting by size.
As side effect directories will appear before files on all filesystem
implementations. To give an example: before this change directories had a size
of 4 KiB when using Linux with ext4 or tmpfs, and with ZFS a size resembling
an estimation of the number of leaves within said directory.
* browse: Catch the case of a directory disappearing before having been read
* browse: Revert to old pass-through behaviour
PROPFIND is a request for an alternate view on a directory's contents, which
response is indeed not implemented but ideally allowed to ask for.
OPTIONS would ideally return (at least) what methods the requestor could use,
which is an allowed request method, too.
This addresses #767.
On matched header rules, replacer is used to replace any placeholders
defined in header rules iex. X-Backend {hostname} where {hostname} will
be replaced by the hostname key present in the replacer
hostname key added to replacer. The value is determined by the output of
`os.Hostname()`
Caddyfile parameter "clients" of "tls" henceforth accepts a special
first modifier. It is one of, and effects:
* request = tls.RequestClientCert
* require = tls.RequireAnyClientCert
* verify_if_given = tls.VerifyClientCertIfGiven
* (none) = tls.RequireAndVerifyClientCert
The use-case for this is as follows: A middleware would serve items to the
public, but if a certificate were given the middleware would permit file
manipulation.
And, in a different plugin such as a forum or blog, not verifying a client
cert would be nice for registration: said blog would subsequently only
compare the SPKI of a client certificate.
As discussed with @mholt I have dropped the old LinkedPath function and
replaced it within the browse template with the new BreadcrumbMap
function. Visually it looks exactly the same as before, now the template
functionality is just more powerful.
Signed-off-by: Thomas Boerger <tboerger@suse.de>
