"Hello World" program core dumps with GHC 4.08.1 on HPUX 10.2
simonmar at microsoft.com
Fri Feb 23 05:16:30 EST 2001
> After much messing about with the gcc and gas sources to help
> track down my
> problems with GHC on HP-UX, I think I've figured why I can't
> built it from
> scratch. I think GCC/GAS must have changed a bit since GHC
> was ported, and
> consequently the evil mangler is being a little too malicious and is
> deleting some ".NSUBSPA" directives that relate to creating
> new subspaces
> and hence the long call fixes that they facilitate.
I'm currently building ghc 4.08.2 on HP-UX using gcc 2.95.2, so we'll
see how that goes. I have one or two local patches in my tree from last
time I did this that may have an impact on the dynamic calling issue.
> My perl skills a decidedly flabby, so I am not sure how
> effectively I can
> fix it, I'll have a look though. I think all it needs to do
> is to keep the
> relevant ".NSUBSPA" line with the function code to which it refers.
> Functions always come out the compiler looking like the example at the
That sounds reasonable: you could try adding NSUBSPA to the T_MOVE_DIRVS
declaration for HPPA in the mangler and see if that makes a difference.
> The other thing I spotted that popped up, but isn't a bug,
> was that I guess
> the "long long" type appeared in GNU C at some point between
> your version
> and 2.95.2, hence the 4.08.1 binary does not include support
> for this and
> due to the configuration procedure in the header files, it
> ends up creating
> a need for this in one of the compiler modules
> I worked around this by copying the libHSrts.a that GHC build process
> produces before it gets onto the compiler from
> ghc-4.08.2/ghc/rts into the
> installed location of the 4.08.1 binary distribution before
> doing a "make
> all" to resume.
Yup, I think I have some local patches to deal with this too. gcc ran
out of registers when compiling some of the long long stuff on HP-UX as
I recall, so I turned it off.
More information about the Glasgow-haskell-bugs