>>>>> "Jonathan" == Jonathan Wakely via Gcc <[email protected]> writes:
Jonathan> That doesn't have to be a problem:
Jonathan> #define ALLOW_VLA \
Jonathan> _Pragma("GCC diagnostic push"); \
Jonathan> _Pragma("GCC diagnostic ignored \"-Wvla\"")
Jonathan> #define DISALLOW_VLA \
Jonathan> _Pragma("GCC diagnostic pop")
Jonathan> ALLOW_VLA;
Jonathan> for (int vla[n]; ; )
Jonathan> ...
Jonathan> DISALLOW_VLA;
Jonathan> That doesn't interfere with the declaration.
A long time ago I filed https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108811
for something like this -- the ability to enable a certain warning for
just a subset of switch statements.
Putting pragmas before and after the switch is possible but it would be
nicer if it could be done with a single line, so instead of
ENABLE_WARNING;
switch (blah) {
// it is gdb so of course there's hundreds of lines here
}
DISABLE_WARNING;
we could write
[[enable_warning_for_this_statement]] switch(...)
I'm indifferent to whether this is an attribute or pragma, it's more
about having it automatically narrow its scope to a single statement.
Jonathan> Somebody has to
Jonathan> develop and test and maintain that attribute forever, for a very
Jonathan> narrow use case which is arguably just user error. Don't use -Wvla if
Jonathan> you want to compile code that uses VLAs.
I'm sympathetic to that but it seems like the machinery is there already
and it's more a matter of how it is applied by the user.
But OTOH maybe the other thing to do is WONTFIX that bug.
thanks,
Tom
