[Mesa-dev] [PATCH] mesa: Allow custom text to be inserted in version string at buildtime
Ian Romanick
idr at freedesktop.org
Wed Apr 24 07:29:00 PDT 2013
On 04/24/2013 04:11 PM, Paul Berry wrote:
> On 23 April 2013 21:22, Chad Versace <chad.versace at linux.intel.com
> <mailto:chad.versace at linux.intel.com>> wrote:
>
> On 04/23/2013 09:19 PM, Eric Anholt wrote:
>
> Chad Versace <chad.versace at linux.intel.com
> <mailto:chad.versace at linux.intel.com>> writes:
>
> On 04/23/2013 06:19 AM, Ian Romanick wrote:
>
> On 04/23/2013 03:28 AM, Chad Versace wrote:
>
> This allows maintainers/packagers/testers to tag the
> build with
> information that will be reported by GL_VERSION.
>
> If the environemt variable or make variable
> MESA_VERSION_STRING_EXTRA is
>
> environment
>
> set, then its values will appear in the GL_VERSION
> string immediately
> after "Mesa X.Y" and before "(git-xxxxxxx)".
>
> This patch implements supports
> MESA_VERSION_STRING_EXTRA only for Android.
> Other build systems are left as an excercise.
>
>
> Why is this useful?
>
>
> The commit message says why: it allows packagers to tag the
> build with info
> that gets reported by GL_VERSION.
>
> How is it used now? The Mesa that gets shipped bi-weekly
> internally in Intel
> for Android is not a snapshot of master nor of any stable
> branch. Yet, that
> snapshot of Mesa is "official", as far as Intel's Android
> efforts are concerned.
> The Android team is using this patch to inject into
> GL_VERSION the name of the bi-weekly
> Android release.
>
> Without this patch, you may one day get a bug report that
> says: "This bug exists
> in Mesa 9.2-devel (git-xxxxxxx)". Then, you will search for
> the git sha1, and
> unable to find it, yell bloody murder and close the bug as
> invalid. With this
> patch, though, you may get a bug report that says: "This bug
> exists in
> Mesa 9.2-devel otc-android-2013-04-23-blah (git-xxxxxxx)",
> and then you
> will know how to find that Mesa.
>
>
> So along with a useless sha1, I have a useless date that doesn't
> tell me
> what Mesa it was actually based off of. I say this having tried to
> figure out what Mesa actually got used in one of these android
> tests and
> failed.
>
>
> Rather than jabs and parrying, let's turn this conversation into a
> constructive one. If you receive a
> communication about Mesa on Android, how do you want the Mesa version
> communicated to you so that you can fetch and then inspect that same
> Mesa?
>
> Keep in mind that the Android Mesa tree is regularly rebased onto
> master, so the shas
> are not as useful as you would hope. Also know that we have git tags for
> the "useless" dates if you know where to fetch from.
>
> Would it be helpful to use "git describe --long" to generate the git
> portion of the GL_version string? That's essentially what the Linux
> kernel build uses it for this kind of purpose. It generates a
> generates a string like this: "mesa-9.1-9-ga5c79b7", where "mesa-9.1" is
> the most recent annotated tag accessible from HEAD, "9" is the number of
> commits between that tag and HEAD, and "a5c79b7" is the SHA of HEAD.
>
> If we do this we might want to start making annotated tags of master
> every now and again--currently the most recent annotated tag accessible
> from master is a tag called "snb-magic" from November 2010.
Regardless of what happens with the rest of this, that is a fine idea.
I'd be in favor of tagging master each time we're going to make a stable
branch. I know GIT can figure out what the common root of two branches,
but it would be a lot easier to just do 'git log pre-9.1..' to see all
the commits since the branch was created.
More information about the mesa-dev
mailing list