Le 12/03/2011 10:49, Lex Trotman a écrit :
On 12 March 2011 20:21, Frank Lanitz frank@frank.uvena.de wrote:
On Sat, 12 Mar 2011 02:53:47 +0100 Colomban Wendling lists.ban@herbesfolles.org wrote:
Maaaybe, sort of see your point, but not really convinced that uprating warnings to errors is a good idea on the dev codebase, it stops people trying and testing things.
Unfortunately, believe me that non-fatal warnings are use to be ignored by unexperienced programmers, believing that if their code compile it is then OK. And I don't see why a warning upgraded to an error on every build would be worst than a syntactical problem (as I described above previously)? In a typical situation, the developer who writes the plugin should get the warning (well, the error), see his plugin don't build, care (hopefully :D), and then fix it directly even before committing and then before anybody else could face the problem. Don't you think?
Just thinking about adding a flag which is activating a paranoid check inside default build system and which lead into failing buildsin such a case. This would give plugin devs the chance to fix their code on one hand and unexperienced people which just want to test to have a build they can do their stuff with.
Yeah, so long as it doesn't make the whole system too complicated and is well documented, it has to be easily usable and maintainable :-)
It would be easy to add a flag to enable/disable the warnings (actually I already had a flag to disable the warnings, thought they was enabled by default). However in Autotools side it's probably only really doable at configure-time, don't know for Waf.
So you'd prefer what: 1) enable warnings by default, and propose some to errors with a flag 2) no warnings by default, can be enabled by a flag (with or wtithout error promotion?)
IMHO 1) is better so the developers gets the warnings, and it would still please everyone since these wouldn't be fatal -- only ugly :D.
If we finally agree with the flags I suggested minus the errors, I could set this up soon. Needs agreement (or not) then :)
Cheers, Colomban