IRC channel logs

<ZhaoM>youpi: I attach gdb on pfinet, returning EINTR in io_select_common happens quite frequently, I supposed it happened only once

<youpi>ZhaoM: take care that pfinet serves not only your test program, but also everything that happens on your system

<youpi>so possibly the EINTR you see correspond to other programs

<youpi>you can set up a second instance of pfinet, and use it thanks to remap or such

<ZhaoM>ok

<ZhaoM>youpi: gdb starts stopping at the break point until

<ZhaoM>youpi: "gdb starts stopping at the break point until I run ./runtest.pl 546", can this be an evidence that no other program is using pfinet?

<youpi>you mean it never stops at the breakpoint before you run ./runtest.pl 546 ?

<ZhaoM>yes

<youpi>then a priori no other program is using pfinet

<youpi>but maybe gdb is

<youpi>iirc it would use pflocal only though

<ZhaoM>it seems debugging in a multi-server system is quite tricky :/

<youpi>try to debug a monolithic kernel ;)

<ZhaoM>I have tried, but didn't go in depth. Maybe it's still tricky :|

<youpi>I mean: debugging a monolithic kernel is way more a problem. You clearly can't just use gdb on that

<youpi>so it's not the fact that it's multi-server that makes it tricky

<youpi>it's the fact that you are debugging the system you are running

<youpi>and actually, being multi-server allows you to escape that

<youpi>as I said, you can set up a second instance of pfinet, and debug that one

<ZhaoM>ok

<Pellescours>what is the interest to have some crypto implementation directly in the kernel? I see that some libs (openssl) can use the kernel implementation (for linux case). Is it for speed? security?

IRC channel logs

2025-02-24.log