hniksic(a)srce.hr said:
"Stephen R. Anderson" <sra(a)bloch.ling.yale.edu>
writes:
> I was somewhat surprised, after posting this failure message,
not
> to get any response at all - not even a flame - from a list that
> otherwise seems rather chatty..... But I interpreted that as an
> implicit rebuke, an injunction to go fix the problem myself.
It's not a rebuke -- it likely means that noone is running
MkLinux
and having a clue about what exactly the problem is here.
Sorry - I meant this as a joke. I guess I should use smileys more liberally.
:-)
Offhand, it appears that your header files don't define `struct
exception'. Grep through /usr/include/**/*.h, find where `struct
exception' is defined, include that, and see if it fixes your
problem. Alternatively, try undefining HAVE_MATHERR and see if that
helps you.
Either way, you should investigate and see why XEmacs configuration
system did the wrong thing with HAVE_MATHERR (`config.log' is a good
place to start checking.)
> So change that "FAILURE" to "SUCCESS" on
MkLinux (egcs-1.02). As
long as it's built with the not-quite-latest development files.
That's not quite the final solution. There is probably a bug in:
1) XEmacs floating point code; 2) XEmacs configuration mechanism; 3)
MkLinux libc headers.
IMHO, the culprit is, in order of decreasing probability, #2, #3 or
#1.
I just received a message from Fred Bacon (<bacon(a)aerodyne.com>), which I hope
he will not object to my quoting:
bacon(a)aerodyne.com said:
I was looking at this very same problem last night. It turns out
that the prototype for matherr has changed slightly. You can get
xemacs to build without any problem by simply changing the variable
on line 934 in floatfns.c from exception to __exception. Everything
else goes smoothly after that. Apparently there used to be a typedef
for exception which has disappeared from the include files.
This implies, unfortunately, that all three of the potential culprits you
suggest are involved in one way or another. I will leave it to those who are
more familiar with the way these things interact in xemacs code to suggest a
patch.
--Steve Anderson