IRC channel logs

<sneek>Welcome back chrislck!!

<rlb>I have no idea why yet, but the powerpc r6rs-ports.test failure is caused by --trs-file, i.e. "./check-guile r6rs-ports.test" is fine, "./check-guile --trs-file foo r6rs-ports.test" fails with the error.

<rlb>ACTION will likely look in to it tomorrow

<sneek>civodul: wb!!

<dsmith>The Horrors!! The Embarrassment!! Guile string->uri-reference does not support fully qualified domain names! See https://debbugs.gnu.org/cgi/bugreport.cgi?bug=53201 for patch with tests

<sneek>Shocking!!

<mwette>sneek so smart

<ArneBab>dsmith: will you be merging it?

<dsmith>ArneBab, I have not such powers

<rlb>civodul: fwiw, I'm still investigating the main r6rs-ports.test failure on powerpc - s390x was fine.

<ArneBab>dsmith: Dale P. Smith is you, right?

<ArneBab>dsmith: does this need a change in docs?

<dsmith>ArneBab, Yes, that's me. Docs? I'll look.

<dsmith>ArneBab, I don't see a need to change the docs.

<ArneBab>Do we have to note that the domain name absolute, if it ends in a dot? "Domain names that end in a dot are called absolute, and are taken as complete." ⇒ https://datatracker.ietf.org/doc/html/rfc1035

<ArneBab>(this is normally only for the files, so not required here, I think)

<ArneBab>it may be good to add a comment in the code ;; fully qualified (absolute) domain name

<ArneBab>Could you also do the small change: gnu commit message format (just check the log)?

<ArneBab>The linked bug actually opened neatly debbugs

<dsmith>Sure

<dsmith>ArneBab, What needs to change in the commit message?

<ArneBab>Just needs an addition below the bug: * path/to/file (procedure name): support fully qualified domain names

<ArneBab>and similar for the test file.

<ArneBab>Just get the git log of the file and adapt another commit :-)

<ArneBab>dsmith: example:

<ArneBab> * libguile/posix.c

<ArneBab> (maybe_warn_about_environ_mutation): New private procedure ...

<ArneBab> (scm_environ): ... called here when mutating the environment ...

<ArneBab> (scm_putenv): ... and here.

<ArneBab> * NEWS: Update.

<rlb>OK, I think I've found the general issue -- those tests use a shared "log" via log!, but I'm fairly certain, don't clear it everywhere they should (only clear when getting results), and the issue is that there are some "late" close calls (presumably from previous port objects) that corrupt the results of later tests.

<dsmith>ArneBab, Hmm. I have this in my local repo as the commit message: https://bpa.st/WWAQ

<ArneBab>dsmith: that looks good — I’ll check whether I looked wrong

<ArneBab>I seem to have imported that wrong

<ArneBab>will check

<ArneBab>dsmith: the patch 2 is missing that part: https://debbugs.gnu.org/cgi/bugreport.cgi?filename=0001-Allow-trailing-.-in-urls.patch;msg=11;bug=53201;att=2

<dsmith>Fouey

<ArneBab>dsmith: did you see my email?

<dsmith>ArneBab, Thanks! Replied with an attachment

<ArneBab>dsmith: looks good to me, the tests pass. Merged and pushed. Thank you!

<dsmith>ArneBab, Cool! Thanks!

<rlb>civodul: pushed a few fixes, including a less ambitious fix for that custom-ports test failures and the addition of -ffat-lto-objects when it's available.

<rlb>With that (and the debian big-endian 32-bit patch we were discussing), powerpc is fine now.

<civodul>rlb: neat!

IRC channel logs

2025-03-02.log