I think you are talking about the case for FAT. we do wait for 2 seconds to exit because the resolution of mtime on FAT is it. though it doesn't help for running many processes to update caches because it isn't a singleton process and other process can access caches and the targeted directories/fonts during updating.
fc-cache might be multi-threading but I'm not sure if writing caches is the thread-safety nor worth doing so. and if we do, doing it as a single process like demonizing may be better, to avoid complication.
On Mon, Nov 28, 2016 at 2:30 PM, Keith Packard <keithp@xxxxxxxxxx> wrote:
Akira TAGOH <akira@xxxxxxxxx> writes:
> The former one was what I took measures for that. I expect it works as long
> as the process is done within the most accurate time.
Is there some place we could add a suitable delay to ensure that file
system without sub-section timestamps could get reliable results using
inexpensive timestamp checking, rather than expensive file scanning? I
recall placing a 'sleep' in the cache generation code, but presumably
that's insufficient?
--
-keith
Akira TAGOH
_______________________________________________ Fontconfig mailing list Fontconfig@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/fontconfig
