la 10.5.2025 klo 11.51 Holger Levsen (hol...@layer-acht.org) kirjoitti: > > control: reassign -1 release-notes > control: tags -1 - moreinfo unreproducible > control: retitle -1 document that rust packages require SSE2 on i386 > thanks > > Hi, > > I'm reassigning this to release-notes as this needs to be documented > for the trixie release. Leaving some context too, but please read the full > bug.
This might indeed be a good idea. IIRC Go packages have similar requirements. > On Sat, May 10, 2025 at 11:37:48AM +0300, Martin-Éric Racine wrote: > > la 10.5.2025 klo 11.23 Holger Levsen (hol...@layer-acht.org) kirjoitti: > > > On Sat, May 10, 2025 at 11:02:55AM +0300, Martin-Éric Racine wrote: > > > > Since the above sqv bin-NMU trickled down to Trixie, APT systematically > > > > fails to verify Debian archive keys: > > > [...] > > > > W: Some index files failed to download. They have been ignored, or old > > > > ones used instead. > > > what are your versions of apt and sqv installed? > > ii apt 3.0.1 i386 > > ii sqv 1.3.0-1+b1 i386 > > > > This only happens on one of my Trixie hosts, so this could be a Rust > > > > issue. > > > > Architecture: i386 (i686) > > > do you have several i386 hosts or only one? what CPU models do they have? > > I have two i386 hosts tracking Testing (currently Trixie). The U1400 > > one updates its APT lists just fine. This Pentium III (Coppermine) has > > been failing to do so ever since sqv 1.3.0-1+b1 trickled into Testing > > yesterday. > > ok, this was to be expected, see #1095862, thus reassigning to release-notes. Debian might as well push i386 over to debian-ports at this point, instead of releasing Trixie with broken support and without kernels, since making Pentium 4 the base CPU level for Rust, Go, etc. will essentially kill the whole x86-32 user base, except for the last few laptop chipsets that were released before AMD and Intel discontinued x86-32 production. > That said, can you please try removing sqv and see what happens? apt should > use > gpgv then instead and that should work. It cannot be removed without --force-depends since APT nowadays has sqv as a hard Depends. It would be a good idea for APT to list supported alternatives e.g. "sqv | gpgv" instead, preferably in the same order as dpkg-dev does. Once sqv was force-removed, APT indeed was able to update its APT lists. However, since a Depends was removed, it now cannot perform any useful operation until that missing dependency is resolved. Martin-Éric
