On Apr 7, 8:19, Daniel Stenberg wrote:
>We're now down to only 2 remaining warnings with the "regular" (non-pedantic)
>icc 8.0 compiler. (and 252 warnings with the pedantic one!)
>Both of these are "warning #269: invalid format string conversion" and they
>both (AFAICS) complain on the use of 'z' as modifier to printf() a size_t
>POSIX 1003.1 says 'z' is the modifier to use:
>... and our internal printf() code supports it.
>It makes me wonder what icc considers is the right choice here? I suspect we
>will have to live with these 2 warnings.
[A rather late reply, I finally got some time to look into some of this.]
AFAIK the 'z' modifier is an OSI C99 feature, and icc is apparently meant to
be a c99 compiler.. it has flags to switch c99 mode on and off etc. (c99 is
default). I haven't found anything in the documented list of incompatibilities.
I have asked on the Intel support forum list if this should be regarded as a
bug in icc. No replies yet, will report later if I get any.
>With my commit yesterday regarding the new auth code, it seems one or two
>tests have started to fail on Tru64. I'm puzzled to why this test fails on one
>particular platform, but I'll try to look over my source code to see if I've
>made any apparent mistakes. Anyone with any suspicitions or clues?
When I tried to reproduce this yesterday, it came out all clear.. so it looked
like whatever was done the last day had fixed the problem. Looking at last
night's builds however, it seems that the 154 test case error is gone, and the
88 one happened on only one of the Tru64 builds. Oh, and on one of the IRIX
I also read today's posting <Pine.LNX.4.58.0404160922050.3576_at_linux3.contactor.se>
which indicates that 88 is actually case 91 all over again.. and it is not
What about that 154 error? It looked fairly consistent (happened on all the
Tru64 builds since before Easter I think), is it fixed? Or just randomly
Anyway, I'll run my autobuild script a bunch of times to see what happens
(in mail-off mode ;-)
Received on 2004-04-16