This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: named warnings & individual warning control

> I don't think you can really do that.  People *will* end up with
> huge source bases with thousands of these #pragmas that are built
> with -Werror (after all, isn't that the point of this feature?)

Um, not really.  The point is that some people, for policy reasons,
have warnings enabled for constructs they are unable to avoid.  If
they end up with thousands of these, something has gone seriously
wrong (or at least, they're using the feature badly).  We don't want
to encourage that.

Also, a change in warning control probably means that something within
gcc changed related to that warning.  It's a valid reason to examine
the warning in the new context and decide if you still need to control

> It needs to be at least as stable as command-line flags are now.

There are about 800 unique messages in the C/MI/MD portions of the
compiler already.  We do not have that many command line options.  For
pragmas that correspond to command line options, sure, we can keep
those reasonably stable.  But the other 750 or so are going to be as
fluid as the warning logic itself.

The reason I think we need to set this expectation is that there was
much concern about locking the gcc developers into the 800 or more
unique mnemonics that will be required for this type of conversion,
especially as many of them will be nonsensical mnemonics (like

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]