Mail Archives: djgpp-workers/2000/12/17/04:41:52
Yesterday tried to build gcc-2.95.3 (current CVS version) for DJGPP.
With some hacking I got it through. Java and Chill related
parts were removed without much thinking:
- I didn't saw any serious interest beyond some poeple
taking not tested binaries from me. No any feedback,
even not in form - they are broken
- Chill support will perhaps be off by default in gcc-3.0
The same will perhaps be with Java
The problems I got was linked with libio. For a rather long time
(before switching to libstdc++-v3) I had problems building development
versions for DJGPP (mostly as cross-compiler) due to broken libio:
- the prototypes of rather many function doesn't fit with
their definitions
This shows only if types of off_t and fpos_t are different (eg. signed
and unsigned ints) and it doesn't appear for example under Linux
and maybe some other systems.
Now this breakage is taken to gcc-2.95 branch. I workarounded it by
patching some files in libio.
Now about other problem I have already mentioned:
I have different problem building gcc-2.95.3 on a new machine I'm normally
using (when in Latvia), it was replaced recently from P200MMX to PIII 700.
There I'm getting SIGSEGV in cc1obj building libobjc. I got it more
than week ago when tried to build gcc-2.95.3 CVS version for DJGPP and
later for Linux (including this morning with the same sources I suceeded
to build it here for DJGPP on a different box)
So maybe there are some problem with that PIII box. It's rather strange
as it sucessfully bootstraps and tests gcc-2.97 CVS version (updating
and tests are run from cron task).
I'ts rather unlikely I'll mess with that before returning to Riga as
serious debugging through SSH connection is not I want to do now.
Perhaps I'll try now to build gcc-2.95.2.1 for Linux and see whether this
problem appears.
Andris
- Raw text -