[uClibc]Hi, upgrading to uClibc 9/6 snapshot ...

Marshall M. Midden m4 at brecis.com
Mon Sep 9 14:54:34 UTC 2002


Original response edited down to not include code/etc from responded to e-mail.

> Some of the link_warnings are there to alert users to the use of unsafe
> functions.  Others are the to alert them to expect differences from
> behavior they might expect.  The locale-related link_warnings fall into
> that category, and are there to cut down on the number of questions
> posted on the list.  If they are made a Config option, the default
> should definitely be "enabled".

Yes, enabled by default!

> As far and the #warning messages go, I suppose I could wrap my #warning
> reminders in #ifdefs.  I really don't understand why you should care
> about them though.

gcc option -Werror is why I care.  I.e. Stop on warnings.  A very good
thing to stop questionable coding.  (we have our kernel compiling with
this too ... linux programers seem to be careless...)   The link_warning's
stop Makefile's if -Werror is applied too.

... re: time.h
> I left these fields in (but disabled) because I will probably implement
> them at some point.   However, they are not ANSI or SUSv3 so I didn't
> include them initially in my time code.  By commenting out the #if 0,
> all you're doing is allowing apps to either get garbage data or
> segfault.

We have time functions implemented -- I grabbed "date" from OpenBSD --
since I put the year 2000 fixes in years ago, I knew it... .

re: uninitialized variables in misc/time/time.c
> -     int oday;                                       /* Note: oday can be uninitialized. */
> +     int oday = 0;                           /* Note: oday can be uninitialized. */
...
> I thought the comments were plain enough... these variable do _not_ need
> to be initialized.  Doing so increases the generated code size on one of
> my target platforms.  So I'm not going to initialize them just to quiet
> some bogus gcc warning.  Maybe I need to add a more detailed comment...

Again, gcc -Werror can't have any warnings.  Now if compiling could
test for -Werror enabled in gcc and do different things... .



More information about the uClibc mailing list