Hello,

Some news on the glibc/hurd debian packages side :)

So glibc 2.34 seems to be alright now after a few fixes.

On the hurd side:

- there was an issue in fakeroot that was making it inexplicably crash
some processes it runs, now fixed.

- I added a fix for terminal reset at end of session. For a long time
now I was noticing that when connecting to a hurd box, sometimes my pty wouldn't be working properly, as if it was left in a bad state by its
previous owner. I noticed that if the master closes and another master
opens it while a slave is still connected, the terminal was not reset.
This is now fixed.

- Last but not least, with glibc 2.34, the call to _hurd_libc_proc_init
by bootstrap processes is working, and thus these processes
(pci-arbiter, rumpdisk, ext2fs) now show up again properly in ps output,
and at shutdown time.

Samuel

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

And also, I have just uploaded version 0.16-3+hurd.of libpciaccess. The
issue was that the map_legacy method was completely broken. Ideally
it'd be able to use the map_region method, but that'd only work for
the proper PCI rom (and even still, since pci-arbiter doesn't get the
rom_base from libpciaccess, we don't know it). It happens that Xorg not
only maps the VGA rom, but also the whole BIOS roms, the video memory,
the interrupt vector table, etc. etc. So I added using the x86 mem_map
as a fallback.

Samuel

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi,

El 23/8/22 a les 19:30, Samuel Thibault ha escrit:

And also, I have just uploaded version 0.16-3+hurd.of libpciaccess. The
issue was that the map_legacy method was completely broken.

What was the problem with it?

Ideally it'd be able to use the map_region method,

Do you mean device_map_region() from the arbiter? If what you want is to
map the ROM, Why not calling the new device_map_rom() method?

What's the context of this?

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Joan Lledó, le mar. 23 août 2022 20:42:20 +0200, a ecrit:

El 23/8/22 a les 19:30, Samuel Thibault ha escrit:

And also, I have just uploaded version 0.16-3+hurd.of libpciaccess. The issue was that the map_legacy method was completely broken.

What was the problem with it?

For a start, it was not specifying a region number, so that was random
and thus opening /server/bus/pci/.../region%d was deemed to almost
always fail :)

Ideally it'd be able to use the map_region method,

Do you mean device_map_region() from the arbiter?

No, I mean pci_device_hurd_map_range, sorry for the mistype.

If what you want is to map the ROM, Why not calling the new
device_map_rom() method?

Because it's not only the rom, but also the vram, other roms, the
interrupt vector table, etc.

What's the context of this?

Xorg, simply :)

Samuel

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)