Ulf Lamping wrote:
Also, the build is failing for me, as "wiretap/file_util.h" includes
"mkstemp.h", but "mkstemp.h" is in the top-level directory, and the
build of Wiretap doesn't do "-I..".
Ok, when I remember this correct, wiretap itself doesn't use mkstemp, so
no need to include it from wiretap.
...but we use g_mkstemp() if present, so it's part of the abstraction
done by wiretap/file_util.h.
I'm not sure where the abstraction of the low-level file access
routines should be done. Wiretap *could* be the right place, although
the stuff it currently handles is a bit higher-level than that;
another possibility would be to have another library that just handles
that.
Another lib *might* be a bit oversized (but I'm not sure).
Yes, that's why I only listed it as "another possibility"; a fairly thin
file-access library might be overkill.
In either case, whatever library handles it should get "mkstemp.c" and
"mkstemp.h", rather than having it be in the top-level directory.
We already have epan/filesystem which *might* fit better (but didn't had
a closer look if it's really the case).
...except that not all programs that use Wiretap use libethereal -
editcap doesn't, for example.
I would like to keep things as smooth as possible for now, as I would
think my changes might introduce some more work to be done.
We might think about mkstemp later (and fix it for now) ...
For now, I'm inclined to go with the "quick fix" and move it (and the
configure script stuff that checks for it) into Wiretap.
