re: Fixing configure failures from newer gcc

[matthew green <mrg%eterna23.net@localhost>]

Martin,

i think you should use the compiler.mk method where necessary,
since it applies to all GCC 14, not just netbsd.

i used CFLAGS.NetBSD hack initially to get as much building as
possible before investigating problems, but the compiler.mk
version is much smarter.

(i will defer to others on this point if the consensus is to
avoid this method.)

> Whether it's broken or not depends on your point of view.  It's good 
> that they are trying to enforce correctness, but breaking software that 
> previously built fine and worked fine for the most part in the default 
> compilation environment is pretty shitty behaviour, and should IMHO only 
> have been done for software that explicitly requested newer standards.

i have pretty conflicted feelings about this, yeah.

on the one hand, all this software was apparently working fine
for decades, so why so much effort now?

on the other hand, i've found a *LOT* of bugs.  because there
are several new classes that are error, i've decided that some
of them are actually important and some are boring.

in particular, the ones i care about most are that missing
prototypes are now an error.  this has led to many failures
as some header isn't included properly, and so we end up with
"int" default for types everywhere.  a lot of these are for
functions with pointers as input or output, so this was
actually a real big in some cases, truncating input (that
in most cases was already 32 bit so no effect, hene why we
did not notice this.)  the other case that comes up is when
functions that don't exist are used and then you end up with
broken plugins.  for instance, some of the "transcode" plugins
have been broken since that package was converted to ffmpeg4,
as it still tries to use functionality from ffmpeg3, and some
of them are using old ImageMagick functions -- there are some
patches to try to make it build against the current version,
but they're incomplete and it won't work.


of the list of ~1200 packages i normally have installed for my
desktop systems, i'm down to just a handful that don't build
anymore.  i've avoided using the compiler warning workaround
as much as possible, but it's there some..


.mrg.