Re: [SLE] Unwelcome warnings from include files
[Philipp Thomas]
This is a bug in gcc 2.95.2 [...] If you want the patch, just send me a mail. [...] IMHO -Wconversion is of limited use [...]
Philipp, let me thank you for the explanation, offer, and advice. In the meantime, I'll merely remove the most cumbersome `-W...', and try restoring some of them after SuSE 7.0.
For the __strsep_g issue you might like to ask on libc-alpha@sourceware.cygnus.com .
I'm not feeling comfortable enough with `libc' development to report a bug which I do not see clearly myself, and I would like to resist diving deep in studying this one. Maybe I'm getting old, maybe things have changed. There are _so_ many things that should work simultaneously, when you prepare a distribution, that one gets easily lost if not trying to stay on track. For example, each time I hit a `libtool', `autoconf' or `automake' problem, that often means _days_ of study, exploration and exchanges with other maintainers, each time. And various `gettext' and `libit' issues are not easy either. [Andreas Jaeger]
Yes, your hope [of "warning-clean" include files] is reasonable - and as glibc developers we strive to fix this but depend on external programs - and not always compile with all possible flags.
I usually try to preset `CFLAGS' to many warnings in the environment at `login' time, and the `CFLAGS' value then gets configured into `config.status' rather automatically. I lost this settings due to some operator error, a while ago, and did not notice. But some of my users have lynx eyes, and got to me with many such accumulated warnings to solve! :-) I guess, Andreas, that `libc' developers could use a similar trick, maybe, to make sure that nothing escape their scrutiny. [Andreas Schwab]
Don't use -Wconversion. It is useless with ISO-C code. It is only usefull if you want to add prototypes to K&R-C code, and want to know about cases where `foo (char)' is actually `foo (int)' in absense of prototypes.
I removed it, Andreas, according to your suggestion and explanations. Thanks to all for your clear and thoughtful replies. P.S. - I usually prefer relating directly to friendly maintainers and individuals, than to wide mailing lists and cold committees. Strangely, the [SLE] list is so friendly that I sometimes have the feeling of writing to an individual. P.P.S. - I have a bit of the same feeling with the [Python] main list, yet it gets diluted as the list seem to quickly grow. I hope that the raising popularity of Python will not kill the spirit, as it did for Perl. Small is beautiful! :-) -- François Pinard http://www.iro.umontreal.ca/~pinard -- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/Support/Doku/FAQ/
François Pinard wrote:
[Andreas Jaeger]
Yes, your hope [of "warning-clean" include files] is reasonable - and as glibc developers we strive to fix this but depend on external programs - and not always compile with all possible flags.
I usually try to preset `CFLAGS' to many warnings in the environment at `login' time, and the `CFLAGS' value then gets configured into `config.status' rather automatically. I would not recommend to do this because this can interfere with CFLAGS settings in many configuration scripts. Imagine to have -mcpu=pentium in a CFLAGS environment and then try to cross-compile a library.
Instead, if you think you need to override CFLAGS, CPPFLAGS etc., better pass them from the command line at configuration time (Should works with properly written autoconf/automake configurations), eg.: CFLAGS=-Wall CPPFLAGS=-DFOO=foo ./configure This essentially is the same as setting them as environment variables, but requires one additional stage of thought, therefore is less error-prone and spares you being hit by invalid CFLAGS settings months later, when you don't recall anymore having set them in the environment. OTOH, you should normally trust in the settings a packages' maintainers think are suiteable. In general they know better than you - but nobody is perfect :) Ralf -- Ralf Corsepius Forschungsinstitut fuer Anwendungsorientierte Wissensverarbeitung (FAW) Helmholtzstr. 16, 89081 Ulm, Germany Tel: +49/731/501-8690 mailto:corsepiu@faw.uni-ulm.de FAX: +49/731/501-999 http://www.faw.uni-ulm.de -- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/Support/Doku/FAQ/
participants (2)
-
corsepiu@faw.uni-ulm.de
-
pinard@iro.umontreal.ca