- Jun 17, 2021
-
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Document the Steam compatibility tool interface See merge request !324
-
Ludovico de Nittis authored
Fix interop with older Upstart- and sysvinit-based OSs See merge request !325
-
Simon McVittie authored
This can be the case on older sysvinit- or Upstart-based Debian-derived distributions, such as Ubuntu 14.04 (with the default Upstart init system) and Debian 9 (with the non-default sysvinit init system). Under systemd or the versions of sysvinit in Debian >= 10, /dev/shm is a real tmpfs and /run/shm is a symlink to /dev/shm. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
bubblewrap mounts filesystems recursively, so mounting /dev is enough, unless /dev/shm is a symbolic link (which will be handled in a subsequent commit). /dev/pts is assumed not to be a symbolic link: there would be no reason to do so. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
This is based on testing with Steam beta 1623823138 and older versions, and is not necessarily authoritative. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
- Jun 15, 2021
-
-
Simon McVittie authored
Update populate-depot.py See merge request !322
-
Simon McVittie authored
CI: Speed up build:devel step See merge request !323
-
Simon McVittie authored
scan-build doesn't understand `__attribute__((cleanup))` (see https://bugs.llvm.org/show_bug.cgi?id=3888 ) so it will report increasingly many false-positive memory leaks in this codebase. It's sufficiently time-consuming that it doesn't make sense to run it as part of our usual pipeline. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
This build pass takes long enough that it makes sense to parallelize. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Now that we're invoking populate-depot.py once per runtime branch, it's straightforward to use different versions for different runtimes. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Since SteamLinuxRuntime v0.20210505.0, the default is to use unpacked runtimes *or* include the archives, but tests/pressure-vessel/containers.py wants both. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
tests/pressure-vessel/containers.py currently expects this, but the default behaviour for populate-depot.py changed between v0.20210430.0 and v0.20210505.0. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
The SteamLinuxRuntime scripts have phased out support for populating a depot directory with a mixture of scout and soldier, because that doesn't really make sense for a game or a compat tool: each native Linux game, and each compat tool used to run non-native games, expects to run in one particular environment (scout or soldier or something else) and using any other environment would be incorrect. However, tests/pressure-vessel/containers.py currently expects to work with a directory that contains both scout and soldier. For now, we'll download scout and soldier separately, then mash them together into the same directory. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
- Jun 10, 2021
-
-
Simon McVittie authored
pressure-vessel: Address several clang warnings See merge request !321
-
Ludovico de Nittis authored
Since the commit fce30b8d, this variable was not used anymore. Signed-off-by:
Ludovico de Nittis <ludovico.denittis@collabora.com>
-
Ludovico de Nittis authored
This commit addresses most of the warnings printed at compilation time while using clang and `ninja scan-build`. Signed-off-by:
Ludovico de Nittis <ludovico.denittis@collabora.com>
-
- Jun 09, 2021
-
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
CI: Set more reasonable permissions in the source package See merge request !318
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Ludovico de Nittis authored
CI: Reduce CI load See merge request !319
-
Ludovico de Nittis authored
Improve compatibility with other mtree implementations See merge request !320
-
- Jun 08, 2021
-
-
Simon McVittie authored
The nochange flag is fully implemented here: it suppresses changing the modification time or permissions. The optional flag is mostly implemented here: it suppresses the error that would ordinarily occur if a file with content does not already exist. The ignore flag is not yet implemented, only parsed, because it only really applies when verifying whether a filesystem tree matches a specification. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
time=1.234 has historically meant 1 second + 234 nanoseconds, or what normal people would write as 1.000000234. Parse it as such, but for now treat this ambiguous format as an error, unless the fractional part is exactly ".0" which is unambiguous - we can relax this to a warning later, if we want to. libarchive and FreeBSD mtree show mtimes with exactly 9 decimal digits, which makes them mean the same thing in either interpretation. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
libarchive allows both, although FreeBSD and NetBSD mtree do not. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
When we push to master it's usually for one of three reasons: 1. We merged a merge request 2. We pushed the "release vX.Y.Z" commit, together with a tag 3. We pushed some other trivial change directly When we merged a merge request, immediately running CI is not necessarily very helpful, because we already ran it for the branch that's getting merged. If we pushed a release commit and a tag, we want the CI pipeline for the tag to take priority over the CI pipeline for the release commit. If we pushed a trivial change, it might not even need CI (if it's just a change to a README, or similar), but if it does, we can trigger it manually. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
If we avoid running pipelines for merge requests, then the status of the pipeline for the branch will show up. The pipeline for the branch can be more comprehensive because it has access to project-specific masked and protected variables. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
To get the deploy job to run, the tags need to be configured as protected so they get to see all the necessary variables. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
It's confusing for it to appear in the job graph when we won't be launching it the majority of the time. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Simon McVittie authored
Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
Ludovico de Nittis authored
Publish pressure-vessel releases, separate from scout releases See merge request !317
-
Simon McVittie authored
When combined with changes to SteamLinuxRuntime scripts, this will let us do a new pressure-vessel release and incorporate it into soldier/sniper/etc. depots, without having to do a whole release of scout to provide the new pressure-vessel. These releases don't include anything that we didn't already publish in Gitlab artifacts, and they're relatively small, so they might as well be public immediately (unlike the full scout releases, which we avoid publishing before they are final, both because they are rather large and because game developers could get compatibility issues if they build against a non-finalized scout runtime). Signed-off-by:
Simon McVittie <smcv@collabora.com>
-
- Jun 07, 2021
-
-
Simon McVittie authored
The production build uses dpkg packages, built with -O2 -g and stripped. This is similar to what we currently extract from a scout SDK build; but if we do it here, we can decouple the pressure-vessel release process from the scout release process, making it possible to issue a new pressure-vessel release for use with pre-existing scout/soldier runtimes much quicker than the time needed to do an entire scout release. Signed-off-by:
Simon McVittie <smcv@collabora.com>
-