Age | Commit message (Collapse) | Author | Files | Lines |
|
Also, rename build-tools/orc.recipe to build-tools/orc-tool.recipe to
match the recipe name. Same as glib-tools.
|
|
The build will fail without this.
https://gitlab.freedesktop.org/gstreamer/gst-editing-services/merge_requests/68
|
|
The corresponding library does not have a pkgconfig file.
|
|
|
|
|
|
The VS solutions do not support Visual Studio 2019 at present, so
fallback to MinGW when building with anything other than 2015 or 2017.
|
|
Jenkins doesn't use --manifest or the `manifest` property. It uses the
`recipes_commits` property to set the commits to be used by Cerbero,
so we should also use that to decide when to disable usage of tarballs
for gstreamer recipes in custom.py
Also, the whole setup was broken for --manifest too because I didn't
test the initial implementation properly.
|
|
|
|
Bumping to 0.50.1 would also work, but it's too close to the 1.16
release to do that.
|
|
|
|
See:
https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/merge_requests/301
https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/merge_requests/182
|
|
This regressed in https://gitlab.freedesktop.org/gstreamer/cerbero/merge_requests/154
|
|
I accidentally pushed the previous incomplete commit straight to master.
|
|
|
|
The debian packaging creates a tarball for each package (such as
base-system-1.0), then extracts it and creates a .deb out of it.
However, package_prefix was being ignored completely so the extracted
tarball was tar-bomb instead of extracting to the package_prefix dir.
Closes https://gitlab.freedesktop.org/gstreamer/cerbero/issues/140
|
|
Fixes the build on macOS.
|
|
|
|
Use a default list of mirrors and add the option
to extend that list in the config.
|
|
Make cURL fail when the download failed, so that instead
of storing the error reply by the server in the destination,
an exception is Raised to fallback to the mirrors
|
|
Closes #138
|
|
|
|
Make sure that autotools projects are configured with PIC
so that they can be linked statically.
This fixes the following linker issues:
dist/linux_x86_64/lib/libkate.a(lib_libkate_la-kate.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libopus.a(opus_multistream_encoder.o): relocation R_X86_64_32S against `.rodata' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libsoup-2.4.a(libsoup_2_4_la-soup-auth.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libwavpack.a(libwavpack_la-common_utils.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libtag.a(id3v2framefactory.cpp.o): relocation R_X86_64_32S against `_ZTVN6TagLib5ID3v212FrameFactoryE' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libspeex.a(speex.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libvorbisenc.a(vorbisenc.o): relocation R_X86_64_32 against `.rodata' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libtheoradec.a(decinfo.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libtiff.a(tif_close.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libjpeg.a(jcapimin.o): relocation R_X86_64_32S against `jpeg_natural_order' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libcairo-gobject.a(libcairo_gobject_la-cairo-gobject-structs.o): relocation R_X86_64_32 against `.bss' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libharfbuzz.a(libharfbuzz_la-hb-buffer.o): relocation R_X86_64_32 against `.rodata' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libfontconfig.a(fccfg.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libexpat.a(xmlparse.o): relocation R_X86_64_32S against `.rodata' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libfreetype.a(ftinit.o): relocation R_X86_64_32 against `tt_driver_class' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libpng16.a(png.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libpixman-1.a(pixman.o): relocation R_X86_64_32S against `.rodata' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libxml2.a(entities.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libgnutls.a(record.o): relocation R_X86_64_32 against `.rodata' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libgmp.a(pprime_p.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a shared object; recompile with -fPIC
dist/linux_x86_64/lib/libtasn1.a(coding.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC
|
|
|
|
|
|
This fixes a crash on Win32. Our old GCC 4.7 isn't smart enough to do it
by itself. On Win32, the stack alignment is 4 bytes, while GCC assumes
16 bytes when using __m128 vectors.
Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/582
Related to https://bugzilla.gnome.org/show_bug.cgi?id=797092
|
|
These four plugins have been generally useful to me, and should be for
other people too: audiolatency, bz2, ipcpipeline, netsim.
We were already enabling the bz2 plugin explicitly, we just weren't
shipping it.
|
|
|
|
In case of static linking a module we also need the .la file
in order to know the dependencies. Create it.
|
|
|
|
Now it is posible to remove a component from an environment variable
|
|
When running under a locale that the system does not support, perl
prints various warnings to stderr which are accidentally captured by
this and lead to an invalid perl version being set.
|
|
This option is only used when the package command will create
a tarball or an Android package (which inherits from DistTarball).
Uses `xz --threads=0` by default to maximize parallelism. This was
compared with `pixz` and found to be either the same or better in
tests.
|
|
Also, always use `tar` on platforms other than Windows. The tar
shipped with MSYS is very old and slow, but it's quite performant on
other platforms.
Speeds up packaging on my machine from 7m30s to 2m3s. The speed up is
even more drastic on the GitLab runners: 2m21s to 3.7s.
Closes https://gitlab.freedesktop.org/gstreamer/cerbero/issues/132
|
|
_create_tarball() does the same thing in both classes.
|
|
This reverts commit d28f934ac6d744d2d6b444709c342ee66250aeff.
Otherwise subprocess invocations will not have the new environment,
which causes a CI failure on macOS during the install phase of this
recipe:
[(72/79) gstreamer-1.0-osx-framework -> install ]
/bin/sh: pkg-config: command not found
/bin/sh: pkg-config: command not found
We can get rid of os.environ manipulation completely in the future.
We already pass the `env` explicitly to most places, but there is
a chance of breakage, f.ex. code that uses `cerbero.ide.pkgconfig`.
So we can only do this in 1.17.x.
|
|
Pointed out by Andoni on IRC:
```
<ylatuya[m]> for example build --missing-files uses ls_files before starting and finally updates it
<ylatuya[m]> this is how it knows what changed before and after the recipe was built
```
|
|
We don't need it, and it causes errors on install due
to lack of permissions because the new meson python
module returns an absolute path into the python install
prefix for
import('python').find_installation().get_path('purelib')
where the old deprecated python3 module returned a path
without a prefix. This is apparently on purpose, so nothing
for us to fix here.
|
|
subprocess.call() outputs to stdout, and that doesn't get
redirected to the log file, which is unnecessarily verbose.
|
|
|
|
Sometimes we call it on the same files multiple times. Speeds up
recipe parsing + packaging from 5.0s to 4.8s.
|
|
The values will always be the same, so cache them. This speeds up
recipe parsing from 5.6s to 5.0s.
|
|
This is massively faster. Speeds up recipe parsing from 8.2s to 5.6s
on Linux.
|
|
|
|
This is faster, especially on Windows, and also fixes these spammy
warnings:
Exception ignored when trying to write to the signal wakeup fd:
BlockingIOError: [Errno 11] Resource temporarily unavailable
|
|
|
|
subprocess.check_call() outputs to stdout, and that doesn't get
redirected to the log file, which is unnecessarily verbose.
|
|
This enabled using one . per MB for progress report. This should greatly
reduce the spam in the CI.
Fixes #131
|
|
|
|
Removes errors such as:
error: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/install_name_tool: input file: /var/folders/tb/6cyjznrj2fb4lq9gd7mdpfm80000gn/T/tmp6jpad8pjlibffi.a is not a Mach-O file
|
|
|