Re: A question about update-excuses (was Re: testing is broken)
email@example.com (David Starner) writes:
> ... but how many packages actually have bugs specific to one architecture?
For what it's worth, my wanna-build databases show 298 failed builds on hppa
and 165 failed builds on ia64 likely to generate bug reports, as soon as I get
time to file them...
Some of these are source packages that are just broken (missing files, scripts
that aren't executable, contents not the expected version, etc). Some are
because hppa is the first architecture using gcc-3.0. Some are because ia64
is "really 64 bits" and authors were lazy about using enough #include
directives to avoid implicit declarations (things that are warnings on a 32
bit arch are often real errors on ia64!).
Almost none of these are *really* architecture specific bugs to my way of
thinking, but often one architecture in Debian gets to trip over them first
and has to take responsibility for getting them fixed... hopefully improving
the overall quality of our code base as a result.