Re: [PATCH] Introduce xcb_aux_alloc_shm() to create anonymous files in RAM
Julien Cristau <[email protected]>
| Newsgroups | gmane.comp.freedesktop.xcb |
|---|---|
| Message-ID | <[email protected]> |
On 04/17/2018 12:52 PM, Alexander Volkov wrote: > 13.04.2018 09:47, Uli Schlachter пишет: >>> +[], >>> +[with_shared_memory_dir=yes]) >>> + >>> +shmdirs="/run/shm /dev/shm /var/tmp /tmp" >>> + >>> +case x"$with_shared_memory_dir" in >>> +xyes) >>> + for dir in $shmdirs; do >>> + case x"$with_shared_memory_dir" in >>> + xyes) >> How many shells are there that do not like "break"? :-/ >> (And does autoconf provide some way to make this loop nicer?) > >> So, if I do --with-shared-memory-dir=/foobar, configure will tell me >> that I specified an invalid directory? Why?? And why does this flag >> exist at all, i.e. what would be uses for it? >> >> Also, why does the existence of a directory at compile time have any >> significance at run time? What about cross compiling? Building a distro >> package with /run/shm existing and then running it on another system >> that does not have /run/shm? >> >> I really think that this compile time check is plain wrong. > > Well, I just copied it from > https://cgit.freedesktop.org/xorg/lib/libxshmfence/tree/configure.ac > The same code is used in the X server: > https://cgit.freedesktop.org/xorg/xserver/tree/configure.ac#n1121 > Somehow it's working there... > What's the use of replicating the code from libxshmfence instead of just calling into that library? Or are they somehow subtly different (and if so why)? Cheers, Julien _______________________________________________ Xcb mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/xcb