Re: [Gnu-arch-users] Linux BKCVS to GNU Arch Script

[Paul Mundt]

On Thu, Feb 03, 2005 at 06:57:15PM +0000, Catalin Marinas wrote:
> My local mirroring proved to be stable for the last ~3700 changesets
> (since Linux 2.6.9). Unfortunately, I don't have a place to host a
> public mirror of it.
> 
Seems to break with current BK, though I have not looked in to it yet:

-- Building the 1.26831 changeset
    Merge bk://bk.arm.linux.org.uk/linux-2.6-rmk into 
ppc970.osdl.org:/home/torvalds/v2.6/linux
1 out of 1 hunk FAILED -- saving rejects to file arch/arm/mm/consistent.c.rej

I haven't had this happen with any previous changesets, but it does seem
to be reproduceable with this one. Have you had any better luck with it?

If we look at the .rej itself, it's obvious why the reject happened, but
it's not clear why the file being tracked has diverged.

***************
*** 293,299 ****
                        vma->vm_flags |= VM_RESERVED;
                        ret = remap_pfn_range(vma, vma->vm_start,
                                              page_to_pfn(c->vm_pages) + off,
-                                             user_size, vma->vm_page_prot);
                }
        }
  
--- 293,300 ----
                        vma->vm_flags |= VM_RESERVED;
                        ret = remap_pfn_range(vma, vma->vm_start,
                                              page_to_pfn(c->vm_pages) + off,
+                                             user_size << PAGE_SHIFT,
+                                             vma->vm_page_prot);
                }
        }
  
If we look at the file itself, the + off seems to be missing from the
page_to_pfn() line, which is what causes the reject. Looking at bk/web,
we see that this was done in the previous revision:

http://linux.bkbits.net:8080/linux-2.5/diffs/arch/arm/mm/address@hidden

but it looks as though this was never applied, even though it clearly
exists in the bkcvs tree too. Looking at the annotation for this:

1.25         (rmk      18-Feb-05):
page_to_pfn(c->vm_pages) + off,
1.26         (rmk      26-Feb-05):
user_size << PAGE_SHIFT,
1.26         (rmk      26-Feb-05):
vma->vm_page_prot);

Yet the tree generated by your script does not appear to reflect that.
If you like, I can send you a diff of the bkcvs tree vs this (as far as
determining if there are other minor breakages or not, I have not looked
in to this yet, but I suspect there may be).

Any ideas?

pgpdJpCA7dHQs.pgp
Description: PGP signature