IRC channel logs

<Pellescours>to compile shadow 4.15.0 we need the constant LOGIN_NAME_MAX to be defined, thing that is not done in hurd. In linux it’s defined in /usr/include/bits/locale_lim.h

<Pellescours>I know that theses files are managed by glibc but I don’t know if this limit should be defined or if somehing like PATH_MAX

<Pellescours>youpi: I’m playing with the flavio’s script cross-hurd and for some reason the patches you pushed this morning make the hurd hang when starting servers (exec startup proc auth). I’m bisecting to find the culprit

<Pellescours>youpi: the culprit is db46ea2eb9dc84959fbf9b1819facac3d6078ba1 when I revert it, it don’t hang

<Pellescours>solid_black: ^

<youpi>ok, reverted it for now

<youpi>thanks for the investigation !

<solid_black>sorry

<solid_black>oh, round_page(-1) breaks things, and I only tested with fresh tasks where there was nothing to deallocate

<solid_black>-vm_page_size then? that the highest number that wouldn't get rounded up to 0

<Pellescours>solid_black: no problem :) I just wanted to notice you about it

<solid_black>thanks for that

<Pellescours>can the other commit that remove does the same thing break things too?

<solid_black>no, since that doesn't do round_page ()

<solid_black>does it break anything for you?

<Pellescours>I don’t think so

<Pellescours>vm_size_t is signed or unsigned?

<solid_black>unsigned

<Pellescours>so round_page( (unsigned) -1) should not round to 0, does it?

<solid_black>round_page rounds *up*

<solid_black>as opposed to trunc_page, which rounds down

<solid_black>so 0xfffffff (many f's), aka -1, gets rounded to 0x0

<Pellescours>ah

<solid_black>0xfffff000 aka -vm_page_size, on the other hand, is the highest number that doesn't get rounded to 0

IRC channel logs

2024-03-23.log