IRC channel logs

<ZhaoM>good afternoon

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

<sneek>ZhaoM, Pellescours says: yes there is a dead lock with rpctrace and gsync_wait, solid_black can probably tell you more about it

<ZhaoM>ok

<crupest>Does Hurd use the same ABI and calling convention as Linux?

<crupest>Abseil can't find symbols of ELF on Hurd.

<crupest>If Hurd does, the symbolization feature should work like on Linux.

<Gooberpatrol66>i don't think it uses the same abi

<Gooberpatrol66>i think you have to recompile programs to the gnu target to get them to work on hurd

<crupest>Sorry, I have said the wrong thing. Not ABI. It is the calling convention. How does it manage stack when calling a function.

<crupest>Abseil has a feature that get symbol name of current function. It uses a very tricky. It first use asm to get PC value and search ELF for sysbol.

<crupest>If the calling convention is the same as Linux, I think it should just work on Hurd.

<Gooberpatrol66>crupest: i'm not sure, here's an article about hurd programming on a lower level, maybe that will be useful https://iandouglasscott.com/2024/02/09/hello-hurd/

<crupest>Nice! I'll read that! Thanks.

<youpi>crupest: it uses the same calling convention as on linux, sysv

<crupest>Interesting, so the symbol searching in abseil should work on Hurd as well. But not.

<crupest>I have to look into it more deeply.

<crupest>Leter.

<crupest>Maybe the issue lies on other places but I miss it.

<crupest>Oh. Now I see why abseil can't find symbols.

<crupest>It uses /proc/self/maps.

<crupest>On Linux, it is 004b7000-004b8000 r--p 00000000 08:01 3278353 /usr/bin/cat

<crupest>While on Hurd, it is 00001000-00002000 r-xp 00000000 00:00 0 [mem_obj=22]

<crupest>No filename, no symbol!

<almuhs>hi. I was testing the latest upgrade of Debian GNU/Hurd in many other Thinkpad models (T60x3, R60e, R61i), and rumpdisk doesn't work in any of this. It seems that rumpdisk only works in the T410, but I don't know the reason

<almuhs>in the other models, rumpdisk continues with the previous errors, losing interrupts and unable to detect correctly the disk (start pixide instead AHCI driver)

<almuhs>added to this, in the T410, the "halt", "poweroff" or "halt-hurd" commands gets an error when try to turn off the wd0. "error reading fsbn 32507920 (...)", after so many "device timeout reading fsbn 32507920"

<almuhs>maybe the rumpdisk problem is related with IOAPIC configuration: the IPI between the device and the cpus are not been sending correctly. And it could be the same reason which cause the failure in CoreDuo/Core2Duo machines: maybe there are different configuration depending of the cpus generation

<almuhs>it's only a idea, not an affirmation

<Pellescours>When doing io intenssive stuff, I get message (default pages): default_write got out of room...

IRC channel logs

2025-02-07.log