IRC channel logs

<rlb>Arsen: ok, I have a patch that tests for and adds -ffat-lto-objects if we already detected -flto. I don't know that it's even possible that you could have one without the other right now, but it was easy to be more selective. Thanks for the help earlier.

<rlb>sneek: later tell civodul don't know if this is suitable upstream, but we still need it in some places - https://salsa.debian.org/rlb/deb-guile/-/blob/812a65543eb9325e220d08e1b00e1a98c6c27936/debian/patches/0007-Fix-32-bit-big-endian-builds-via-Oresolve-primitives.patch (double-checking powerpc right now)

<sneek>Got it.

<civodul>sneek later tell rlb hi! re https://salsa.debian.org/rlb/deb-guile/-/blob/812a65543eb9325e220d08e1b00e1a98c6c27936/debian/patches/0007-Fix-32-bit-big-endian-builds-via-Oresolve-primitives.patch i think there would first need to be a bug report and then we could discuss a fix/workaround

<sneek>Will do.

<civodul>sneek: botsnack

<sneek>:)

<civodul>ACTION wonders how much botsnack sneek gets on a day

<sneek>Welcome back dsmith :D

<dsmith>sneek, botsnack

<sneek>:)

<chrislck>sneek: botsnack

<sneek>:)

<mwette>sneek: botsnack

<sneek>:)

<rlb>civodul: there is a bug, see line 13 - https://salsa.debian.org/rlb/deb-guile/-/blob/812a65543eb9325e220d08e1b00e1a98c6c27936/debian/patches/0007-Fix-32-bit-big-endian-builds-via-Oresolve-primitives.patch

<sneek>Welcome back rlb, you have 1 message!

<sneek>rlb, civodul says: hi! re https://salsa.debian.org/rlb/deb-guile/-/blob/812a65543eb9325e220d08e1b00e1a98c6c27936/debian/patches/0007-Fix-32-bit-big-endian-builds-via-Oresolve-primitives.patch i think there would first need to be a bug report and then we could discuss a fix/workaround

<rlb>There's also discussion in the debian bug.

<Arsen>rlb: no problem

<Arsen>I'd propose removing the flto configury (to be clear, I am a big fan of LTO, and guile should encourage it, but best let the user actually add the flag)

<Arsen>if such a discussion happens, please CCC me

<Arsen>CC*

<rlb>Understood, for now I was just fixing the more immediate issue.

<Arsen>yes, reasonable :)

<rlb>...I still get regular "inter-protocol attack" failures in the repl-server test when running in parallel.

<rlb>(on various hosts)

<civodul>rlb: oh sorry, i hadn’t seen the reference

<civodul>i also get frequent failures of the “inter-protocol attack” test

<rlb>No worries - I should have called it out (not all patches have all the right metadata).

<civodul>ouch, a 2020 bug

<rlb>So far main's being a bit cranky on powerpc -- with that patch and the one to stop test-out-of-memory from being a failure, and ignoring the inter-protocol failures, it's *almost* passing the tests.

<rlb>fwiw

<rlb>And powerpc (for debian) is not a critical arch -- it was just the 32-bit big-endian one I remembered to try first.

<rlb>It could be we need some more of the debian patches there, or powerpc/unstable is unstable right now :)

<rlb>Definitely be nice to figure out what's going on with the inter-protocol failures, though, or mark them as unresolved *if* that's not a security issue. I haven't investigated myself there yet, and don't know anything about the test.

<rlb>(Oh, and fwiw, most of the higher numbered debian patches there are cherry-picks from main that'll be obviated by a release.)

<rlb>civodul: I'm also seeing (a newish I think?) failure there in r6rs-ports.test

<rlb>i.e. "ERROR: r6rs-ports.test: custom textual ports: output port: writing 2000 a's - arguments: ..."

<rlb>Wondering if it may have something to do with the test harness envt. I still see it without parallelism, but not if I run the test separately via "./check-guile r6rs-ports.test"...

<rlb>Yeah, that test just doesn't work on powerpc atm -- I'll try s390x next (also big endian I think, though not 32-bit).

IRC channel logs

2025-03-01.log