Fontconfig - Feb 2006 - [PATCH] never try to validate cache files when using fc-cache -f

Hi, 

2006-02-02  Dirk Mueller  <dmueller@suse.com>

	* fc-cache/fc-cache.c (scanDirs): Don''t check for
	cache validity if force is in effect. 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: make_fccache_f_work.diff
Type: text/x-diff
Size: 635 bytes
Desc: not available
Url :
http://lists.freedesktop.org/archives/fontconfig/attachments/20060202/9ae51c68/make_fccache_f_work.bin

Dirk Mueller wrote:
> Hi, 
> 
> 2006-02-02  Dirk Mueller  <dmueller@suse.com>
> 
> 	* fc-cache/fc-cache.c (scanDirs): Don''t check for
> 	cache validity if force is in effect. 
This patch doesn''t do what you''re saying.  The effect of this
patch is
to always blow away cache files when -f is given, and that''s going to
be
problematic on multi-arch installations.  The check a couple of lines above:

	if (!force && FcDirCacheValid (dir) && FcDirCacheHasCurrentArch
(dir))

checks for cache validity and ought to ensure that -f will try to save
the cache files.  Of course it might fail for some other reason.

-f works for me.  Does it not work for you?  What should the semantics
of -f be?  Perhaps there should be -f and --really-force, which would
always remove cache files; that''s certainly not what we want to give
when we make install.

pat

On Thursday 02 February 2006 18:47, Patrick Lam wrote:
> 	if (!force && FcDirCacheValid (dir) &&
FcDirCacheHasCurrentArch (dir))
>
> checks for cache validity and ought to ensure that -f will try to save
> the cache files.  Of course it might fail for some other reason.
Hmm, indeed. but I want one switch to make sure no corrupt cache files stay 
around, as we have way too many problems with them. 

The man page says: 

   Force  re-generation of apparently up-to-date cache files, overriding the 
timestamp checking.
> -f works for me.  Does it not work for you?  What should the semantics
> of -f be?  Perhaps there should be -f and --really-force, which would
> always remove cache files; that''s certainly not what we want to
give
> when we make install.
what does make install have to do with it (curious)? Indeed, the normal 
behaviour whould be to just run "fc-cache" I think. 


Dirk

Dirk Mueller wrote:
> On Thursday 02 February 2006 18:47, Patrick Lam wrote:
> 
> 
>>	if (!force && FcDirCacheValid (dir) &&
FcDirCacheHasCurrentArch (dir))
>>
>>checks for cache validity and ought to ensure that -f will try to save
>>the cache files.  Of course it might fail for some other reason.
> 
> Hmm, indeed. but I want one switch to make sure no corrupt cache files stay
> around, as we have way too many problems with them. 
It would be great to have a --really-force flag, then.  The man page
doesn''t think about the issues involved in multiarch cache files.
>>-f works for me.  Does it not work for you?  What should the semantics
>>of -f be?  Perhaps there should be -f and --really-force, which would
>>always remove cache files; that''s certainly not what we want to
give
>>when we make install.
> 
> 
> what does make install have to do with it (curious)? Indeed, the normal 
> behaviour whould be to just run "fc-cache" I think. 
make install runs fc-cache -f -v, as I''d imagine postinst scripts would
as well.

pat

Le jeudi 02 f?vrier 2006 ? 14:50 -0500, Patrick Lam a ?crit
:
> Dirk Mueller wrote:
> > On Thursday 02 February 2006 18:47, Patrick Lam wrote:
> > 
> > 
> >>	if (!force && FcDirCacheValid (dir) &&
FcDirCacheHasCurrentArch (dir))
> >>
> >>checks for cache validity and ought to ensure that -f will try to
save
> >>the cache files.  Of course it might fail for some other reason.
> > 
> > Hmm, indeed. but I want one switch to make sure no corrupt cache files
stay
> > around, as we have way too many problems with them. 
> 
> It would be great to have a --really-force flag, then.  The man page
> doesn''t think about the issues involved in multiarch cache files.
> 
> >>-f works for me.  Does it not work for you?  What should the
semantics
> >>of -f be?  Perhaps there should be -f and --really-force, which
would
> >>always remove cache files; that''s certainly not what we
want to give
> >>when we make install.
> > 
> > 
> > what does make install have to do with it (curious)? Indeed, the
normal
> > behaviour whould be to just run "fc-cache" I think. 
> 
> make install runs fc-cache -f -v, as I''d imagine postinst scripts
would
> as well.
Well, I''m doing fc-cache -f in %post of fontconfig package but older
version of caches are sometime in the way and only rm
-f /var/cache/fontconfig/* can fix it :(

-- 
Frederic Crozat <fcrozat@mandriva.com>
Mandriva

Frederic Crozat <fcrozat@mandriva.com> ????????:
> Le jeudi 02 f?vrier 2006 ? 14:50 -0500, Patrick Lam a ?crit :
>> Dirk Mueller wrote:
>> > On Thursday 02 February 2006 18:47, Patrick Lam wrote:
>> > 
>> > 
>> >>	if (!force && FcDirCacheValid (dir) &&
FcDirCacheHasCurrentArch (dir))
>> >>
>> >>checks for cache validity and ought to ensure that -f will try
to save
>> >>the cache files.  Of course it might fail for some other
reason.
>> > 
>> > Hmm, indeed. but I want one switch to make sure no corrupt cache
files stay
>> > around, as we have way too many problems with them. 
>> 
>> It would be great to have a --really-force flag, then.  The man page
>> doesn''t think about the issues involved in multiarch cache
files.
By the way, another option to remove cache files in
/var/cache/fontconfig for directories which don''t exist anymore would
be nice. Deleting such cache files always is probably not a good idea
because directories might be temporarily unavailable (e.g. nfs
mounts). But having an option to clean up old junk in
/var/cache/fontconfig without having to delete everything there would
be nice.

-- 
Mike FABIAN   <mfabian@suse.de>   http://www.suse.de/~mfabian
?????????????