https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=3500
--- Comment #28 from Michael Mann <mmann78@xxxxxxxxxxxx> 2011-06-16 06:03:26 PDT ---
(In reply to comment #27)
> In randpkt.c, the use of ssize_t is, not entirely surprisingly, inside "#ifndef
> _WIN32"/"#endif". It's only used in code that uses /dev/random, which is a
> UN*Xism, and the value is ssize_t because that's what read() returns on UN*X.
> gtk/export_object.c is a bit more work, but we could probably add a ws_ssize_t
> type in wsutil/file_util.h and have it be ssize_t on UN*X and int on Windows,
> to match the return value of write() on UN*X and _write() on Windows.
> That should let us get rid of the definition of ssize_t.
Unfortunately, ssize_t is more prevalent in wireshark-win*-libs, so removing
the definition from config.h (via config.h.win32) after cleaning up
wsutil/file_util.h and gtk/export_object.c doesn't work.
This leaves a few choices:
1. Make config.h.win32 have a definition similar to pyconfig.h (which I can
submit as a patch once I have it working).
2. Create a config.h.win64 file and have the ssize_t definition be different
than config.h.win32 (Again, a patch I can contribute)
3. Modify the source in wireshark-win*-libs to be independent of the ssize_t
definition in config.h. It appears there was a conscience effort to
specifically use config.h, so I don't know if that should be undone. Since this
is not part of the SVN and I'm an outsider, I wasn't sure how that would be
accomplished (if prefered)
--
Configure bugmail: https://bugs.wireshark.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
