Am 13.09.18 um 23:22 schrieb Chris Lamb:
> Nice work. Is this…
>
>
> https://salsa.debian.org/systemd-team/systemd/commit/2e03e24734ef51cd47e1dd0ded3a701bbc936cbf
Those are the changes from
https://github.com/systemd/systemd/pull/10067
which drop the usage of the hard-coded
Hey Michael,
> Thanky you all for your input. I think we found a solution upstream
> which works reasonably nice, so I've cherry-picked those changes and
> they seem to be sufficient to make the the build reproducible.
Nice work. Is this…
Am 13.09.18 um 09:51 schrieb Chris Lamb:
> Hi Vagrant,
>
>> Is it plausible to strip the embedded build directory information out of
>> the test binaries with the binaries shipped in the package
Just curious: which tool would you use for such a case?
> But why not hardcode said path to
Hi Vagrant,
> Is it plausible to strip the embedded build directory information out of
> the test binaries with the binaries shipped in the package, since they
> will almost certainly not have access to the original build directory,
> and it interferes with the reproducibility of the test
On 2018-09-10, Michael Biebl wrote:
> Am 10.09.18 um 19:22 schrieb Chris Lamb:
>>> Now, I could try and convince upstream to not embed ABS_BUILD_DIR into
>>> the binary and maybe let the build system pass the build dir as env var.
>>> This would make those test-binaries a bit more cumbersome to
Am 10.09.18 um 19:22 schrieb Chris Lamb:
> Hi Michael,
>
>> Now, I could try and convince upstream to not embed ABS_BUILD_DIR into
>> the binary and maybe let the build system pass the build dir as env var.
>> This would make those test-binaries a bit more cumbersome to use though,
>> because
Hi Michael,
> Now, I could try and convince upstream to not embed ABS_BUILD_DIR into
> the binary and maybe let the build system pass the build dir as env var.
> This would make those test-binaries a bit more cumbersome to use though,
> because you'd have to manually set the env var if you
Am 10.09.18 um 04:28 schrieb Michael Biebl:
> Thanks Chris
>
> Am 09.09.18 um 21:31 schrieb Chris Lamb:
>> Spent a couple moments on this. The first thing that can be addressed
>> is the absolute CATALOG_DIR which ends up in the /usr/lib/systemd/
>> tests/test-catalog binary in the systemd-tests
control: severity -1 minor
thanks
On Sun, Sep 09, 2018 at 08:28:14AM +0200, Michael Biebl wrote:
> Severity: wishlist
I've upgraded to minor because a.) it's a violation of a "should"
directive in policy (thus wishlist is too low) and b.) systemd is
only unreproducible with variation of the
Thanks Chris
Am 09.09.18 um 21:31 schrieb Chris Lamb:
> Spent a couple moments on this. The first thing that can be addressed
> is the absolute CATALOG_DIR which ends up in the /usr/lib/systemd/
> tests/test-catalog binary in the systemd-tests package.
>
> This dir is currently "useless" in that
Chris Lamb wrote:
> (Sorry to be a downer but I wouldn't want to give the impression this
> was so "easily" remedied. Alas.)
Spent a couple moments on this. The first thing that can be addressed
is the absolute CATALOG_DIR which ends up in the /usr/lib/systemd/
tests/test-catalog binary in the
Dear Michael et al.,
> > ok, FTBR with build path variation.. so the workaround is easy
^^
> > this smells like a simple thing coming from some toolchain
Ah, I see. Well, given that the
Am 09.09.18 um 11:49 schrieb Chris Lamb:
> Hi Michael,
>
>> Talking about that on #debian-systemd, h01ger mentioned that this might
>> not be too hard.
>
> Ah, interesting; did Holger have any first steps that would suggest
> this? A very quick glance at the diffoscope output and the build logs
Hi Michael,
> Talking about that on #debian-systemd, h01ger mentioned that this might
> not be too hard.
Ah, interesting; did Holger have any first steps that would suggest
this? A very quick glance at the diffoscope output and the build logs
(you appear to obey our extra reproducibility related
Source: systemd
Version: 239-8
Severity: wishlist
User: reproducible-bui...@lists.alioth.debian.org
Usertags: buildpath
I notice that systemd fails to build reproducibly [1].
Given the importance of the package I think it would be great if we
could fix that.
Talking about that on #debian-systemd,
15 matches
Mail list logo
