IRC channel logs

<damo22>if SIGTRAP is blocked from the signal mask, and i am getting a bunch of mach_msg_trap calls simultaneously in all threads and a hang, should i add SIGTRAP to the allowed signals?

<damo22>will that stop the process hanging?

<Time2Dot0Beta>What hardware functionality does Hurd use for its RPC context switches? (I’m looking into how the fastest equivalent Linux can do is apparently eventfd based on normal syscall context switches, and the fastest of all, UIPI (user interrupts) are only available on some Xeon server CPUs).

<damo22>itanium64 had one instruction

<damo22>there is no such equivalent for x86

<damo22>there is swapgs

<damo22>thats about it

<Time2Dot0Beta>damo: Yes, there is. Sapphire Rapids has UIPI: https://www.felixcloutier.com/x86/senduipi

<Time2Dot0Beta>(Okay, not pure x86, as an extended instruction set.)

<damo22>ok never heard of that

<damo22>x86 is pretty dead freedom-wise to me

<youpi>sneek: hello

<sneek>Welcome back youpi, you have 2 messages!

<sneek>youpi, damo22 says: sorry, i submitted v3 for smbfs without testing, so there is a 7th one-liner patch to fix it up, i did test that

<sneek>youpi, damo22 says: __pthread_timedblock send a mach_msg without setting MACH_RCV_INTERRUPT, therefore is uninterruptible until the timeout expires: #1 0x0000000001087a0d in __GI___mach_msg (msg=msg@entry=0x3fc2cc0, option=option@entry=0x102, send_size=send_size@entry=0x0, rcv_size=rcv_size@entry=0x20, rcv_name=0x30, timeout=0x32, notify=0x0) at ./mach/msg.c:111

<damo22>oops old ones

<youpi>I know :) just flushing them

<damo22>wb

<damo22>i am compiling your change now

<damo22>i almost didnt see your patch so i wrote my own version as well, but im testing yours

<damo22>mine was slightly wrong :\

<damo22>the handy thing about a goto is that you dont need to indent any code to add a loop :D

<damo22>your test program with the bug shows waiting, then handle forever on my current system (no changes)

<damo22>but i think it is supposed to show waiting then handle only a few times, then waited?

<youpi>damo22: yes

<damo22>i will test with the change once it compiles.. its doing test suite :|

<youpi>you can use DEB_BUILD_OPTIONS=nocheck to avoid the testsuite

<youpi>you can re-run with -nc to avoid the rebuild

<damo22>hmm tempting

<damo22>this crten

<damo22>dh_installman: warning: Multiple definitions for manpage debian/locales/usr/share/man/man8/locale-gen.8 via different compressions. Please ensure there is at most one definition.

<damo22>dh_installman: error: This feature was removed in compat 13.

<youpi>yes, there is an issue with multiple builds, you can just rm -fr debian/*/usr/share/man

<damo22>youpi: the bug is gone from your test program using your glibc patch, but java still hangs

IRC channel logs

2026-08-23.log