Age | Commit message (Collapse) | Author | Files | Lines |
|
Not that it matters at this point.
|
|
GNU Make 4.3 has a backwards incompatible change affecting the use of
number signs or hashes (ie., #) inside function invocations. See:
https://lists.gnu.org/archive/html/info-gnu/2020-01/msg00004.html
In this case, it would expand the '\#' in the '\n\#include \"$(h)\"'
argument to the foreach call to '\#', not '#'. This would lead to
spurious backslashes in front of the '#include' directives in the
generated fs-enumtypes.c file.
Spotted by Ernestas Kulik.
https://gitlab.freedesktop.org/gstreamer/common/-/merge_requests/4
|
|
|
|
http to https, Bugzilla to GitLab, autotools to meson.
|
|
This is a follow up to https://gitlab.freedesktop.org/gstreamer/gstreamer/merge_requests/109
|
|
Need to ignore leak of system-wide once allocated memory.
|
|
|
|
This is needed to get the export define the GST_*_API markers
map to when compiling gst libs.
https://bugzilla.gnome.org/show_bug.cgi?id=797185
|
|
This will be needed soon by the GST_*_API decorators.
|
|
Instead of ignoring a bazillion slightly different stacktraces,
we can just ignore problem where the ld lib is involved, because
we can't control the way it will or will not free its resources
(see
http://pubs.opengroup.org/onlinepubs/009696699/functions/dlclose.html)
|
|
Filesystem listing doesn't guarantee a stable order.
https://bugzilla.gnome.org/show_bug.cgi?id=794069
|
|
|
|
'cp -u' is not specified in POSIX and it can fail with illegal option
error on systems not supporting it. It is better to fallback to the
slower 'cp' than causing build failure.
https://bugzilla.gnome.org/show_bug.cgi?id=791784
|
|
We were still using the old name ffmpeg and thus the entirety of
ffmpeg was being downloaded every time the script was run.
|
|
Adds the necessary xsl templates for tracer's based entirely on the
existing element template
https://bugzilla.gnome.org/show_bug.cgi?id=791253
|
|
Unsurprisingly, runs a test in gdb until it fails somehow (or gdb does).
https://bugzilla.gnome.org/show_bug.cgi?id=790791
|
|
gtk-doc 1.26 was converted to Python and the file was removed.
Copy the needed function and remove the require gtkdoc-common.pl
https://bugzilla.gnome.org/show_bug.cgi?id=786361
|
|
This avoids the following -Wportability warning from automake:
warning: orcc_v_gen_$(V: non-POSIX recursive variable expansion.
Based on the fix in https://bugzilla.gnome.org/show_bug.cgi?id=705869
|
|
So we can e.g. add GST_EXPORT in front of our enum _get_type() functions.
|
|
|
|
We're only interested in updates that update the actual
translations, so ignore changes that just update the date.
|
|
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=781145
|
|
It's no longer needed or supported as docs are in gst-docs now.
|
|
For thread-safety reason we purposefully leak the list of custom
log handlers
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=774641
|
|
We should not make the fact that we do not find a pluginsdir
fatal as for gst-plugins-good we have no way to retrieve
it but still it is needed in the uninstalled case.
|
|
And add getpluginsdir here as it is shared by many modules
|
|
This reverts commit b18d820635185eb549530af1ce406937141b2dd9.
Causes problems for distros with /bin/sh = dash
|
|
Preserve multiword arguments that are to be passed
to configure, and quote them in the generated
autoregen.sh
|
|
That wasn't meant to go in, it's for all modules after all.
|
|
These are system and configuration dependent.
|
|
|
|
Recently noticed that gtk-doc stopped to extract the internal
documentation (appeared just API, but no description).
The problem seems to be a regression from commit f363b32, where the
expasion of the variable _source_dir is expanded using parentheses,
which are used for command substitution, when the variable need to
be a string expansion, and for that brace are needed by the shell.
https://bugzilla.gnome.org/show_bug.cgi?id=768571
|
|
As some places we where assuming a single directory. This caused some
plugins documentation to be ignored.
|
|
unused errors on osx.
On osx, make check in gstreamer directory currently outputs:
"error: argument unused during compilation: '-pthread'".
This new ax_pthread.m4 adds -Qunused-arguments when needed.
Source:
http://www.gnu.org/software/autoconf-archive/ax_pthread.html
History:
http://git.savannah.gnu.org/gitweb/?p=autoconf-archive.git;a=history;f=m4/ax_pthread.m4
https://bugzilla.gnome.org/show_bug.cgi?id=747954
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=763764
|
|
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=761718
|
|
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=762707
|
|
The exported symbols will depend on the backends built
and be different from platform to platform. Should probably
be the win32 symbols anyway, which is not the case at the
moment. So for now just show any diff but not fail.
|
|
|
|
Set it as part of the AM_TESTS_ENVIRONMENT variable so that it
is the same when running make-check from the top-level dir or
when running the element.check variations
https://bugzilla.gnome.org/show_bug.cgi?id=761472
|
|
|
|
On certain filesystems like FAT, NTFS and some network file systems,
symlinks are not supported. In those case, installing the pre-commit
hook fails and leaves the developers with having to run the style
checker manually.
Instead, fallback to copy. This should enable the commit hook for
users of filesystems and certain OSs where symlinks are not supported.
https://bugzilla.gnome.org/show_bug.cgi?id=759089
|
|
the 2-parameter invocation:
AG_GST_SET_PACKAGE_RELEASE_DATETIME([yes], [YYYY-MM-DD])
is greping the first parameter instead of the second.
https://bugzilla.gnome.org/show_bug.cgi?id=739079
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=739079
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=739079
|
|
Something is not right with these checks, they sometimes lead
to build failures on OSX/iOS like e.g.
Undefined symbols for architecture i386:
"rrwarn=%all,no%E_EMPTY_DECLARATION,no%E_STATEMENT_NOT_REACHED,no%E_ARGUEMENT_MISMATCH,no%E_MACRO_REDEFINED,no%E_LOOP_NOT_ENTERED_AT_TOP", referenced from:
implicit entry/start for main executable
ld: symbol(s) not found for architecture i386
Let's just remove the -Wall/-Werror support in case
of this compiler which is most likely completely unused
anyway these days. Code will still compile just fine as
before (hopefully), it's just that we don't get extra
warnings.
|
|
|
|
GST_ENABLE_EXTRA_CHECKS #define
Assertions are useful to have in any case, and we were abusing the #define for
adding extra checks to our code. Let's do it properly!
https://bugzilla.gnome.org/show_bug.cgi?id=756870
|