Thomas Kahle wrote:
> Hi everyone,
> 
> i'm trying to write an ebuild and I experience problems with ccache
> (only on this single package).
> It seems as ccache somehow generates false positives on this one. After
> changing the make options the program fails to build ("Declaration of
> function not found"-like errors).
> Deactivating ccache solves the problem.
> I searched a little in Bugzilla, and there are some bugs where people
> are arguing that ccache can or cannot create problems. So let me phrase
> some clear questions:
> **) Are there known issues with ccache? It seems to be stable and
> untouched for several years already.

None that I personally know of.  It may be the package's build system does 
something insane.  You may want to mail the ccache list with the specifics. 
They're generally responsive.

> **) How can I filter the usage of ccache from inside the ebuild ?

ccache is enabled by FEATURES, and we really can't mess with FEATURES from 
inside an ebuild because it's portage-specific (other package managers may 
handle it differently or not support ccache at all).

> **) Has anyone ever tried to use the same cache for 2 different versions
> of the gcc ("share the cache")? Is it safe?

ccache caches its data with a hash which is composed of, among other things, the 
compiler executable's size and mtime.  when you switch to another version of 
gcc, the mtime and size are different and so the cached data isn't used; in fact 
  the data is cached again.  you can switch back and forth between versions with 
gcc-config and ccache will only use the cached data created with that version.

> I first tried to share the cache, ran into the beforementioned trouble
> and thought its because of the sharing. But after clearing the cache and
> experimenting a little, I found that it seems to be related just to the
> package. Its reproducible with a single compiler.
> 
> Any comments appreciated.
> Thanks
> Thomas

Hope you figure it out.


-- 
fonts,                                            by design, by neglect
gcc-porting,                              for a fact or just for effect
wxwindows @ gentoo     EFFD 380E 047A 4B51 D2BD C64F 8AA8 8346 F9A4 0662