Enabling -ffile-prefix-map by default

Vagrant Cascadian vagrant at reproducible-builds.org
Wed Jun 17 06:03:24 BST 2020


On 2020-06-16, Vagrant Cascadian wrote:
> On 2020-06-13, Holger Levsen wrote:
>> On Fri, Jun 12, 2020 at 02:27:50PM -0700, Vagrant Cascadian wrote:
> We've got some updated numbers now!
>
> unstable (with reproducible=+all):
>
>   15 FTBFS packages in unstable/amd64:
>
>      analitza firmware-microbit-micropython grantlee5 ignition-common
>      kalarmcal kcodecs kdeclarative kookbook kparts libkgapi
>      massif-visualizer okteta scram seqan tellico
>
> bullseye (with reproducible=+all,-fixfilepath after some builds have
> been resceduled):
>
>   5 FTBFS packages in bullseye/amd64:
>
>     grantlee5 ignition-common kalarmcal kcodecs tellico
>
> Some of those (e.g. firmware-microbit-micropython) are not in bullseye
> at all.
>
> I've rescheduled a few of them again, just to be sure they were built
> with -fixfileprefix in bullseye.
>
> Many of them appear to go from FTBFS to reproducible, FWIW.
>
> So we're looking at something in the order of 10 packages that FTBFS
> because of this build flag alone that we've identified.
>
> The "knotifications" package is tagged with this issue but builds
> reproducibly in all cases, so it's possible it includes a fix or
> workaround that might also be applicable for the others?

Nothing obvious from a quick look; "knotifications" was one of the
original packages where this issue was reported.


We also have another relevent issue with several packages tagged:

  https://tests.reproducible-builds.org/debian/issues/bullseye/ffile_prefix_map_passed_to_clang_issue.html

  4 FTBFS packages in bullseye/amd64:

    chromium llvm-toolchain-8 llvm-toolchain-9 llvm-toolchain-6.0#

vs

  11 FTBFS packages in unstable/amd64:

    aiscm android-platform-art chromium goxel httping libblocksruntime
    llvm-toolchain-10 llvm-toolchain-8 llvm-toolchain-9 thunderbird
    wasi-libc

The packages that seem to have been "fixed" were rebuilt recently (maybe
Holger noticed this issue and triggered rebuilds for those too?). The
packages that are still failing on bullseye appear to FTBFS most of the
time on tests.reproducible-builds.org (e.g. llvm-*).


So I'd guess we are looking at more like ~20 identified packages with
issues related to this build flag... and quite possibly some that have
not been categorized... obvioiusly not sure on the margin of error for
the unknowns.


live well,
  vagrant
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 227 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/reproducible-builds/attachments/20200616/d3486ae4/attachment.sig>


More information about the Reproducible-builds mailing list