Guy Harris wrote:
On Nov 17, 2003, at 12:26 PM, Ulf Lamping wrote:
If you didn't get a warning about "dlg_set_cancel()" in
"gtk/capture_info_dlg.c", that must be because MSVC++ doesn't warn about
functions with no prototype in scope - you have to include "dlg_utils.h"
to get that declared.
If you didn't get warnings about "struct pcap_stat *" being used in the
declaration of "capture()" without "struct pcap_stat" being defined,
that's probably also because MSVC++ doesn't warn about that.
I think, I have to install a linux machine (presumably a Debian
distro), and try to compile things, before committing to the CVS tree.
That might help somewhat, by adding more warnings - but it wouldn't have
helped with the initialization checkin; somebody using the IRIX MIPS
compiler had a problem with that one.
Hmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm.
After having a look at the current MSVC warning configuration, I know
why this happens.
A short note about the MSVC warnings:
MSVC knows about 5 warning levels: 0 (nothing) - 4 (very verbose)
As I'm writing portable code regularly, warning level 3 is a good
tradeoff between too much warnings, and don't see the "important" ones.
So I had a look at the warning level that our current configuration uses
, but only the default warning level is used. After looking at the MSVC
documentation, the default warning level is set to 1, so "nearly
nothing" will be warned!!!
After setting the warning level to 3, and recompiled *only* the gtk
specific things, I got about 170 new warnings!!!!
I'm in doubt, if we should raise the warning level to 3, as this will
give us *lot's* of warnings in a win32 generation (didn't tested this,
but this must be thousands).
I really would like to have a clean build environment (this will make
committing code much easier).
So it might be a good idea to reduce current warnings (about 250 in the
current CVS, with warning level 1!), raise warning level by one, reduce
warnings again, until we reach warning level 3 with a reasonable number
of warnings.
Regards, ULFL
