When parsing commits and tags cgit made too many assumptions about the
formatting of said objects. This patch tries to make the code be more
prepared to handle 'malformed' objects.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
I run an instance of lighttpd for cgit behind nginx (nginx
doesn't execute CGI). So the port (SERVER_PORT=33333) that
lighttpd runs on sends to cgit is different from the standard
port 80 that public clients connect to (via nginx).
This was causing the Atom feed URL to show the private port
number that lighttpd was running on.
Since the HTTP/1.1 "Host" header includes the port number if
running on a non-standard port, it allows non-client-facing HTTP
servers to transparently generate public URLs that clients can
see.
So use the "Host" header if it is available and fall back to
SERVER_NAME/SERVER_PORT for some clients that don't set
HTTP_HOST.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
The callback from read_tree_recursive just needs to check the type of
each tree entry; if it's a dir we want to continue scanning, if it's a
regular file we'll assume it's the one we requested.
And while at it, remove some stray fprintfs.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
In commit a1266edfe the build instructions for the git libs where moved
to their real targets, which in turn depended on the phony target `git`.
But since `git` is an actual directory in cgit the git libs wouldn't be
recompiled when needed.
So with this patch (third time lucky), cgit is declared to depend on the
really phony target `libgit` and the build instructions for `libgit` is
to unconditionally rebuild git/libgit.a and git/xdiff/lib.a.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
The urls for plain view makes it possible to download blobs without knowing
their SHA1, but the function needs to be promoted and the link from tree
view seems like a perfect fit.
PS: Although hidden, the blob view still is nice for direct blob access so
there's no point in removing it.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
This patch implements basic support for cloning over http, based on the
work on git-http-backend by Shawn O. Pearce.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
This enables a page which generates atom feeds for the current branch and
path, heavily inspired by the atom-support in gitweb.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
When `local-time` is set, commit, tag and patch timestamps will be printed
in the servers timezone. Also, regardless of the value of `local-time`,
these timestamps will now always show the timezone.
Signed-off-by: Stefan Naewe <stefan.naewe@atlas-elektronik.com>
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
When building cgit we depend on xdiff/lib.a and libgit.a in the git
directory, but the previous attempt on describing this dependency
failed since the build instructions for the libs was placed under the
phony `git` target.
This patch fixes the issue by moving the build instructions to their
real targets. It also makes it clear that only the `cgit` target
depends on the git binaries (since they're only used during linking).
And while at it, the patch also cleans up the list of phony targets.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
The default max-length used when printing commit messages and repo
descriptions can be increased due to the new layout (no sidebar).
Also, on the repo summary page I believe it makes sense to only show the
ten most recent branches and tags by default, just as it makes sense to
show the ten most recent commit messages for the active branch.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
If either repo.clone-url or clone-prefix is specified in cgitrc, all
space-separated values in the config option is printed as a possible
clone url on the repo summary page.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
The objectfiles depends unconditionally on some specific git binaries while
those git binaries depends on the phony `git` target and this patch seems to
get these dependencies spelled out correctly.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
In GIT-1.6.0, read_tree_recursive takes an extra void pointer for callback
data. We might want to use this to avoid some global variables, but for now
lets just make sure that we can still compile.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
The new cgitrc option `footer` can be used to include a html-file which
replaces the standard 'generated by cgit' message at the bottom of each
page.
Suggested-by: Peter Danenberg <pcd@wikitex.org>
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
If blob is invoked with no id=, it tries to look up h= and search for path= in
there. Once found, proceed as normal, otherwise, fail as normal.
Signed-off-by: Michael Krelin <hacker@klever.net>
When an object in the tree has GITLINK mode-bits we don't need to get any
more info about that particular object (and trying to get more info about
it will usually generate an annoying warning on stderr since the object
typically doesn't exist in the repo anyways).
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
The change to print_slot() in cdc6b2f8e7 made
the function return correct errno for read errors while ignoring write errors,
which is not what was intended. This patch tries to rectify things.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
If print_slot() fails, the client will be served an inferior response.
This patch makes sure that such an error will be returned to main(), which
in turn will try to inform about the error in the response itself.
The error is also printed to the cache_log, i.e. stderr, which will make
the error message appear in error_log (atleast when httpd==apache).
Noticed-by: Jim Meyering <jim@meyering.net>
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
These functions handles EINTR/EAGAIN errors during read/write operations,
which is something cache.c didn't.
While at it, fix a bug in print_slot() where errors during reading from the
cache slot might go by unnoticed.
Noticed-by: Jim Meyering <jim@meyering.net>
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
We'll need proper return-values from these functions to make the cache
behave correctly (which includes giving proper error messages).
Noticed-by: Jim Meyering <jim@meyering.net>
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Both cgit_print_diff() and cgit_diff_tree() handles root commits nicely,
but cgit_print_commit() forgot to check the case of 0 parents.
This fixes it, and adds tests to avoid future regressions.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
This makes it more obvious how to get back to the index, especially when the
config option `logo-link` is used. And the page name displayed in the header
provided no extra information. It only consumed space and deserved to die.
While at it, make sure that the different parts of the header doesn't wrap
when horizontal space is limited.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
