| 1 |
On Thu, Dec 29, 2005 at 10:35:12AM +0000, João Brázio wrote: |
| 2 |
> On 12/29/05, Brian Harring <[1]ferringb@g.o> wrote: |
| 3 |
> > On Thu, Dec 29, 2005 at 10:24:06AM +0000, Jo?o Br?zio wrote: |
| 4 |
> > > Wel.. I've already tryed to chroot() into the grp-x86-20051228 and |
| 5 |
> > > issued: |
| 6 |
> > > # emerge portage |
| 7 |
> > > Calculating dependencies |
| 8 |
> > > |
| 9 |
> > > !!! Problem in sys-apps/portage dependencies. |
| 10 |
> > > !!! [Errno 38] Function not implemented: |
| 11 |
> > > '/var/cache/edb/dep//usr/portage/sys-apps/.update.23778.portaege-2.0.53 |
| 12 |
> > > ' exceptions |
| 13 |
> > utime or rename offhand... |
| 14 |
> Excuse me but what do you mean with "offhand" ? |
| 15 |
|
| 16 |
That name for a file is only created with a flat_list cache backend, |
| 17 |
specifically when it's doing an update to an existing entry (kind of a |
| 18 |
duh there considering the name, I know). |
| 19 |
|
| 20 |
The algo is roughly |
| 21 |
f=open(tmp_update_entry) |
| 22 |
write to it |
| 23 |
close it |
| 24 |
utime it (reset mtime) |
| 25 |
rename(tmp_update_entry, update_entry) |
| 26 |
|
| 27 |
Hence the 'offhand'. Don't know if it's rename or utime that's not |
| 28 |
defined- just know that those are the only two syscalls that could |
| 29 |
sanely trigger that (failed update will trigger an unlink, but I'd be |
| 30 |
amazed if that call was missing). |
| 31 |
|
| 32 |
~harring |
Archives