Bug 10020

Created attachment 8770 [details] [review]
patch for m32r

Sorry about the phenomenal bug spam, guys.  Adding xorg-team@ to the QA contact so bugs don't get lost in future.

That patch appears to be against a monolithic X server, which means the odds of this getting applied are astonishingly low.  Is there a corresponding patch for a modular X server?

I am not the original bug reporter, I forwarded your note to him.

It seems to me that this patch applies to current git:
* xorg/util/imake (first hunk, applies fine)
* xorg/util/makedepend (second hunk, applies fine)
* xorg/util/cf (other hunks, probably with an offset)

From my point of view, this patch is against the modular tree. I agree it changes some utility programs (imake and makedepend) that were used for building the monolithic tree. I don't know whether these programs are still used by the modular server nowadays. But, if the submitter wants these utility programs to work on m32r, and if they are still maintained, the patch should be applied. If they are not maintained anymore, I don't see the point of keeping them in the repo, especially if you don't want to receive any other bug report like this one :)

Thanks.
Brice

Created attachment 9174 [details] [review]
moduler X-server

hi,
the elf loader is no longer used, as we just use libdl instead, so that part of the patch isn't necessary.
your pci patch also looks flat-out wrong to me: defining it to use linuxPciInit and then ifdef'ing the entire thing out is a bit ... weird.  if m32r has no pci bus at all, then either you should define your own no-op pci init function; or you can just use the linux one, and it will try /proc/bus/pci, and fail because of that.

> the elf loader is no longer used, as we just use libdl instead, so that part of
> the patch isn't necessary.
Is it meaning elf loader's patch should be removed when using libdl?

> your pci patch also looks flat-out wrong to me: defining it to use linuxPciInit
> and then ifdef'ing the entire thing out is a bit ... weird.  if m32r has no pci
> bus at all, then either you should define your own no-op pci init function; or
> you can just use the linux one, and it will try /proc/bus/pci, and fail because
> of that.
Sorry, the patch for Pci.c was including workaround.

Thanks!
Kazuhiro Inaoka

hi,

(In reply to comment #7)
> > the elf loader is no longer used, as we just use libdl instead, so that part of
> > the patch isn't necessary.
> Is it meaning elf loader's patch should be removed when using libdl?

yes, that's correct.  that file doesn't even exist in the current server.  but as long as libdl works on m32r, then you're okay.

> > your pci patch also looks flat-out wrong to me: defining it to use linuxPciInit
> > and then ifdef'ing the entire thing out is a bit ... weird.  if m32r has no pci
> > bus at all, then either you should define your own no-op pci init function; or
> > you can just use the linux one, and it will try /proc/bus/pci, and fail because
> > of that.
> Sorry, the patch for Pci.c was including workaround.

okay, so we can just ignore this part of the diff?

cheers.

(In reply to comment #8)
> hi,
> (In reply to comment #7)
> > > the elf loader is no longer used, as we just use libdl instead, so that part of
> > > the patch isn't necessary.
> > Is it meaning elf loader's patch should be removed when using libdl?
> yes, that's correct.  that file doesn't even exist in the current server.  but
> as long as libdl works on m32r, then you're okay.
Yes, our libdl works. Please ignore them.

> > Sorry, the patch for Pci.c was including workaround.
> okay, so we can just ignore this part of the diff?
Of course.

Thanks
Kazuhiro Inaoka

I've merged your patch, minus the xf86sym bits (as they would seem to be indicative of other problems), the Pci.c patch, and the elfloader patch.  To work, configure.ac would still need to be changed; please send an updated patch that works with the autotools when you update to current X sources.

Removing keyword and block.

Removing keyword and block.

imake is dead.  Do you still care about this landing?

I don't know if M32R is still relevant, but the X server basically no longer has any per-arch #defines. As of 1.18 the code this patch touched no longer exists.