On Tue, Oct 30, 2018 at 7:05 AM Keith Packard <keithp@xxxxxxxxxx> wrote: > Hrm. I didn't understand what the uuid files were being used for. That is to identify font directories and cache files identifically even if those are *bind-mounted*. > Reading through the code, it looks like they are an attempt to share > cache files when there is more than one path name used for the same > directory? And that by storing the UUID file in the font directory, you > provide a stable name for the cache file even if the directory is > accessed through a different path name. Correct. > If this is true, then I'm not sure I can see how we could move the .uuid > files to a separate directory and retain this property. Yes. there was some idea to implement this but it was simplest way to do. > Is my understanding of why this mechanism was created correct? If so, > I'd like to suggest that we just remove it; I think it's more important > for fontconfig to avoid modifying the font directories than for it to > avoid replicating cache information in the (rare?) case that there is > more the one path to the same font directory. Not rare today. this has been made to improve flatpaks startup time because regenerating fontconfig caches always happened. > We could use the 'realpath(3)' function during the generation of the > cache file name; that would at least avoid having symlinks cause > problems. That won't help in this case. this isn't introduced by symlinks at all. > > -- > -keith -- Akira TAGOH _______________________________________________ Fontconfig mailing list Fontconfig@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/fontconfig
