IRC channel logs

2020-03-10.log

back to list of logs

***Server sets mode: +nt
***me is now known as Guest32776
***me is now known as Guest86572
***me is now known as Guest77730
<janneke>any idea why make 4.3 does not work/hangs? i just built it on the latest vm and anything besides --version or --help hangs after a while with (rpctrace)
<janneke>... => task192(pid22798)->gsync_wait (19578944 2 0 0 0)
<youpi>janneke: I'm not aving such issue
<youpi>ah, 4.3, sorry
<youpi>I don't have this version
<janneke>yeah, guix uses v4.3, it's kind of involved to use another version for the hurd
<youpi>did you try to build & run make 4.3 on debian?
<youpi>just to make sure
***kreyren is now known as kreyren-mail
***kreyren-mail is now known as kreyren
<janneke>i'm trying to bisect git now, i report if/when i have found the offending code
<janneke>yes, that's what i'm doing now
<youpi>k
<janneke>i'm running plain upstream (not make-dfsg), but there are no code changes by theh looks of it
<janneke>haha, found it: 749a54d7a458dc6779936138caf40ce600a80052 is the first bad commit
<janneke>look, this is the commit message
<janneke> * job.c (child_execute_job): Prefer posix_spawn() over fork()/exec()
<youpi>oh
<youpi>posix_spawn is normally tested relatively well by libc
<youpi>but possibly there are still bugs
<janneke>yeah it's ironic that fork/exec works better :-)
<janneke>it's a pretty big patch, though (200LOC); could be something "around" posix_spawn
<janneke> http://git.savannah.gnu.org/cgit/make.git/commit/?id=749a54d7a458dc6779936138caf40ce600a80052
<youpi>posix_spawn is really not a trivial thing actually
<youpi>it potentially does much more than just fork/exec
<youpi>since one can open/close files, chdir, etc.
<janneke>yes, after reading fork last year i was naively hoping to find a much easier spawn :)
<youpi>it's true that a pure spawn is much simpler than fork+exec
<youpi>since you bascially only have the exec part and almost nothing of fork
<youpi>but there is a lot more in spawn than just fork+exec :)
<janneke>hmm, yes i would probably only need a simple spawn (or a simple fork+exec, no threads etc)
<janneke>so...i have a workaround for guix' (bootstrap)make, now i'm wondering what to do with this "bug"
<janneke>"someone" interested an capable with make and the hurd/glibc would have to look at it, some time, i guess