Johan Herland [Wed, 9 Jun 2010 23:09:35 +0000 (01:09 +0200)]
ui-commit: Preserve path limit in links to commit page
This includes adding a path argument to cgit_commit_link() and updating all
its callers. The callers from within the commit page (i.e. the "commit",
"unidiff"/"side-by-side diff" and "parent" links) all preserve the path
limit of the current commit page. All other callers pass NULL (i.e. no path
limit).
Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
The path breadcrumb navigation at the top of the 'tree' page has now been
duplicated in ui-shared, which leaves the ui-tree implementation unnecessary.
This patch removes the breadcrumb navigation from ui-tree, and moves the
"(plain)" link that followed the breadcrumb when displaying blobs to the
end of the next line, following the blob SHA1.
Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Johan Herland [Wed, 9 Jun 2010 23:09:29 +0000 (01:09 +0200)]
ui-shared: Teach "breadcrumb" navigation to path limit display beneath tab bar
When a path limit is in effect, and displayed directly beneath the tab bar,
it should offer breadcrumb navigation (like what the 'tree' page does), to
allow changing the path limit easily.
Implementing this requires a robust way to link back to the current page with
a changed ctx->qry.path, but without losing track of the other query
arguments. This is solved by adding the new cgit_self_link() function, which
is then invoked repeatedly by the new cgit_print_path_crumbs() function while
manipulating ctx->qry.path.
Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Johan Herland [Wed, 9 Jun 2010 23:09:27 +0000 (01:09 +0200)]
ui-shared: Replace ctx.qry.path with ctx.qry.vpath
In all cases where ui-shared uses ctx.qry.path, it is done so in the context
of a in-project path, and not in the context in which the 'refs' page or the
'clone'-related functionality uses ctx.qry.path. Make this explicit by using
ctx.qry.vpath instead.
This path introduces no fundamental difference in functionality except for
fixing some minor bugs, for example the Atom feed reference from a
"$repo/refs/heads/" page.
Note that the usage of ctx.qry.path in the other ui-<page>.c files is ok,
since that code presumably is only executed in the context of its own <page>,
so the correct interpretation of ctx.qry.path is never in question.
Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Johan Herland [Wed, 9 Jun 2010 23:09:26 +0000 (01:09 +0200)]
struct cgit_cmd: Differentiate between various usages of ctx.qry.path
For many commands/pages (e.g. 'tree', 'diff', 'plain', etc.), the
ctx.qry.path argument is interpreted as a path within the "virtual" project
directory structure. However, for some other commands (notably 'refs', and
the clone-related commands) ctx.qry.path is used in a different context (as
a more or less "real" path within the '.git' directory).
This patch differentiates between these two usages of ctx.qry.path, by
introducing a new variable - ctx.qry.vpath - which is equal to ctx.qry.path
in the former case, and NULL in the latter.
This will become useful in future patches when we want various pages and the
links between them to preserve existing in-project paths.
Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Johan Herland [Wed, 9 Jun 2010 23:09:25 +0000 (01:09 +0200)]
ui-shared: Remove needless cgit_get_cmd() call, and refactor hc() accordingly
The call to cgit_get_cmd() and the following fallback handling (to "summary"
or "repoindex") in cgit_print_pageheader() is unnecessary, since the same
fallback handling was already done when ctx.qry.page was set when
cgit_get_cmd() was called from process_request() in cgit.c.
As such, hc() can also be rewritten to simply compare the given 'page' string
against ctx.qry.page.
Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Mark Lodato [Tue, 9 Feb 2010 15:12:43 +0000 (10:12 -0500)]
html: properly percent-escape URLs
The only valid characters for a URL are unreserved characters
a-zA-Z0-9_-.~ and the reserved characters !*'();:@&=+$,/?%#[] , as per
RFC 3986. Everything else must be escaped. Additionally, the # and
? always have special meaning, and the &, =, and + have special meaning
in a query string, so they too must be escaped. To make this easier,
a table of escapes is now used so that we do not have to call fmt() for
each character; if the entry is 0, no escaping is needed.
Mark Lodato [Tue, 9 Feb 2010 04:04:41 +0000 (23:04 -0500)]
html: make all strings 'const char *'
None of the html_* functions modify their argument, so they can all be
'const char *' instead of a simple 'char *'. This removes the need to
cast (or copy) when trying to print a const string.
Georg Lukas [Sat, 28 Nov 2009 02:44:33 +0000 (03:44 +0100)]
"max-blob-size" config var to limit generated HTML size
Sometimes it is not feasible to generate the HTML pretty-print for large
files, especially if a source-filter is involved or binary data is to be
displayed. The "max-blob-size" config var allows to disable HTML output
for blobs bigger than X KBytes. Plain downloads are not affected.
Georg Lukas [Thu, 19 Nov 2009 11:14:45 +0000 (12:14 +0100)]
syntax highlighting for all formats supported by "highlight"
The highlight tool can be given any of the supported file extensions
as its -S parameter. This patch replaces the case-switch by extracting
the extension from the supplied file name and passing it to highlight.
However, this requires a shell supporting the ${var##pattern} syntax,
like dash or bash.
Unknown extensions cause a fall-back to plain text using the --force
switch. Error messages are redirected to /dev/null.
A special case maps Makefile and Makefile.* to the "mk" extension.
The total overhead is reduced by calling "exec highlight". No forks are
needed during script execution.
In side-by-side diff, add support for marking individual characters.
Refuses to do so if the left hand side of the diff has different amount of
differing lines to the right hand side to avoid confusion.
Note that I use the naive dynamic programming approach for calculating the
longest common subsequence. We could probably be more efficient by using a
better algorithm. The LCS calculating function is O(n*m) and uses up n*m
amount of memory too (so if we we compare two strings of length 100, I use
an array of 10000 for calculating the LCS). Might want to not calculate LCS
if the length of the line is too large.
Fixed side-by-side diff bugs related to binary diff and more.
The fixed bugs:
* "Binary files differ" did not show up either in unidiff or
side-by-side-diff.
* Subproject diffs did not work for side-by-side diffs.
* The ssdiff link on diff pages did not conserve the path.
Signed-off-by: Ragnar Ouchterlony <ragnar@lysator.liu.se> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Lars Hjemli [Sun, 6 Sep 2009 17:33:10 +0000 (19:33 +0200)]
ui-plain.c: only return the blob with the specified path
When a path to a directory was specified for the 'plain'
view, each blob in the directory used to be returned to
the client. This patch fixes the issue by matching the
path of each blob against the requested path.
Noticed-by: Lars Stoltenow <penma@penma.de> Signed-off-by: Lars Hjemli <larsh@slackbox.hjemli.net>
Lars Hjemli [Mon, 24 Aug 2009 11:31:49 +0000 (13:31 +0200)]
cgit.c: respect repo-local 'snapshots' option for --scan-path
The repo-specific 'snapshots' option is bitwise AND'ed with the global
'snapshots' option during parsing, and since the global cgitrc hasn't
been parsed when --scan-path is processed the global 'snapshots' will
always be 0 (i.e. no repo-specific 'snapshots' setting will have any
effect).
This patch fixes the issue by setting the global 'snapshots' mask to
0xFF (hence relying on later parsing of the generated cgitrc repolist
to do the right thing).
Lars Hjemli [Mon, 24 Aug 2009 11:27:15 +0000 (13:27 +0200)]
cgit.c: only print first line of repo.desc in print_repo()
Since repo.desc might have been populated by reading the 'description'
file in GIT_DIR, it may contain newlines. And by printing the literal
value, we may then generate an invalid cgitrc include-file.
Lars Hjemli [Mon, 24 Aug 2009 08:22:21 +0000 (10:22 +0200)]
cgit.c: add missing options to print_repo()
Note: print_repo() still ignores repo.max-stats and repo.snapshots,
which both requires additional work since these settings are represented
internally as an enum and a bitmap.
Lars Hjemli [Sun, 23 Aug 2009 22:04:58 +0000 (00:04 +0200)]
Add support for repo-local cgitrc file
When recursively scanning a directory tree looking for git repositories,
cgit will now parse cgitrc files found within such repositories.
The repo-specific config files can include any repo-specific options
except 'repo.url' and 'repo.path'. Also, in such config files the 'repo.'
prefix can not be used, i.e. the valid options then becomes:
* name
* clone-url
* desc
* ower
* defbranch
* snapshots
* enable-log-filecount
* enable-log-linecount
* max-stats
* module-link
* section
* about-filter
* commit-filter
* source-filter
* readme
Lars Hjemli [Sun, 23 Aug 2009 21:23:20 +0000 (23:23 +0200)]
cgit.c: refactor repo_config() from config_cb()
The new function repo_config() is used to handle all 'simple' repo
options, for the following reasons:
* code readability
* parser performance
* upcoming support for repo-local cgitrc files during scanning
Lars Hjemli [Sun, 23 Aug 2009 21:09:31 +0000 (23:09 +0200)]
ui-repolist.c: sort by section name, repo name as default
When no sorting is requested by the client, cgit will now sort by
section name followed by repo name. This allows repos to be registered/
discovered independently of their display order.
Lars Hjemli [Fri, 21 Aug 2009 12:26:52 +0000 (14:26 +0200)]
Rename "linenumbers" to "enable-tree-linenumbers", change default to "1"
This makes the name of the cgitrc option more descriptive and at the
same time changes the default from "0" to "1" in an attempt to stay
backwards compatible - prior to the introduction of "source-filter"
and "linenumbers", cgit always generated linenumber links in the
tree view, but now this feature can be turned off (one might want to
do this if the source-filter performs line-wrapping etc).
While at it, the documentation is updated to match the surrounding
descriptions.
Lars Hjemli [Thu, 20 Aug 2009 08:07:51 +0000 (10:07 +0200)]
cgit.c: fix caching keyed on PATH_INFO with no QUERY_STRING
When generating a hash for caching, ctx.qry.raw is used as key. And
since cgit_parse_url() zero-terminates it's argument (after the repo
path), ctx.qry.raw must xstrdup(ctx.qry.url).
Lars Hjemli [Mon, 17 Aug 2009 07:05:13 +0000 (09:05 +0200)]
cgit.c: do not segfault on unexpected query-string format
The querystring_cb() function will be invoked with a NULL value when
the querystring contains a name not followed by a '='. Such a value
used to cause a segfault, which this patch fixes.
Lars Hjemli [Mon, 17 Aug 2009 07:19:05 +0000 (09:19 +0200)]
ui-tag.c: do not segfault when id is missing from query-string
The purpose of the tag page is to print info about a specific tag,
but if no tag was specified on the query-string cgit used to segfault.
With this patch, cgit will fallback to the value of the 'h' parameter
instead (which is never NULL due to prepare_repo_cmd() in cgit.c).
It will now also verify that the specified tagname is in fact a valid
ref in the 'refs/tags/' namespace, i.e. specifying 'id=master' will
trigger a 'Bad tag reference' error.
Noticed-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Stefan Bühler [Sun, 16 Aug 2009 17:35:18 +0000 (19:35 +0200)]
ui-refs.c: improve handling of lightweight tags
When a lightweight tag is referencing a commit object, cgit now uses
the commit date when comparing tag age. Also, the commitdate and author
info is printed in the refs view, making lightweight tags appear similar
to annotated tags.
Signed-off-by: Stefan Bühler <lighttpd@stbuehler.de> Signed-off-by: Lars Hjemli <hjemli@gmail.com>
git.ckatri.com / cgit.git / log