g++ generates an internal compiler error when attempting to compile the very
first file in GNU/OBST. The source file in question is 11023 lines in
length, and the compiler error occurs on line 11023 (which is the closing
scope of a 1000 statement extern “C” function).
I am really interested in getting my hands on a persistant object store for
RtP. If anyone knows of one (free or commercial) I would appreciate a
pointer.
g++ generates an internal compiler error when attempting to compile the very
first file in GNU/OBST. The source file in question is 11023 lines in
length, and the compiler error occurs on line 11023 (which is the closing
scope of a 1000 statement extern “C” function).
I am really interested in getting my hands on a persistant object store for
RtP. If anyone knows of one (free or commercial) I would appreciate a
pointer.
btw: Who are “they”, I mean, presumably, obst compiles with g++ on linux
(at least they configure for linux and don’t mention anything in the readme
about an internal error with g++ ? Surely, this is unique to the
version of g++ that RtP is using, is it not ?
btw: Who are “they”, I mean, presumably, obst compiles with g++ on linux
‘They’ are the gcc developers - a lot of people from Cygnus and Code Sourcery
and a whole other bunch of volunteers from around the galaxy.
(at least they configure for linux and don’t mention anything in the readme
about an internal error with g++ > > ? Surely, this is unique to the
version of g++ that RtP is using, is it not ?
Well, yes and no. There are sometimes ICE’s that only occur under RTP.
Some of them tend to do with the way we manage (or mis-manage) memory,
others are related to host system dependent portions of the gcc code.
You say it builds under linux? What version of g++?
cburgess@qnx.com> > wrote in message news:94535k$6ju$> 1@nntp.qnx.com> …
Yes, g++ is fairly buggy sometimes. > > They are working on a
maintenance release.
In the meantime, try fiddling with the optimisation level. Turning
off optimisation for that file would probably do it.
Well, yes and no. There are sometimes ICE’s that only occur under RTP.
Some of them tend to do with the way we manage (or mis-manage) memory,
others are related to host system dependent portions of the gcc code.
Yeah, I get the feeling this one is related to RtP paging implementation; I
mean
1000 statements is a lot for a single function, but it is hardly something
that wouldn’t
be encountered fairly regularly in the Linux world. In this case, the
function in
question was simply 1000 function calls (uh, I mean method invocations),
no branching, no looping, no local variables, nothing to optimize really.
You say it builds under linux? What version of g++?
A much earlier version (this code is circa 1995); in fact, I have almost got
it built, and
the biggest problem I encountered is that the authors used non-ANSI (since
ANSI
C++ didn’t exist then) scoping rules throughout the code, and the latent
conformist in
me, beckons me to fix that rubbish