Where is regerror() in MINIX-ST 1.5.10.3 ?
Darius S. Naqvi
I managed to compile "less" last night. I used the version on
ftp.Uni-Koeln.DE (134.95.80.1). You have to define REGCOMP (not REGCMP)
to be 1 in defines.h in order for it to compile. But even then,
_regerror was an undefined symbol. It turns out the the library
functions regcomp() and and regsub() call regerror(), which isn't
defined anywhere. As a quick fix, I put a line like
#define regerror(p) fprintf(stderr, p)
in /usr/src/lib/other/regexp.c and /usr/src/lib/other/regsub.c, compiled
them, and put them in libc.a in place of the old regexp.o and regsub.o.
Then "less" linked fine, and ran once I chmem'd it to 32000.
Am I missing some part of the library? I'm sure that regerror() is
meant to be defined somewhere, but I never saw it anywhere. Is this
fixed in patch 4 for 1.5.10 (i.e. 1.5.10.4) for the ST?
--
Darius S. Naqvi mail:dar...@edm.isac.ca
ISA Corp. phone:(403) 420-8081
Edmonton, Alberta, Canada
Michael Haardt
> I managed to compile "less" last night. I used the version on
> ftp.Uni-Koeln.DE (134.95.80.1). You have to define REGCOMP (not REGCMP)
> to be 1 in defines.h in order for it to compile. But even then,
> _regerror was an undefined symbol. It turns out the the library
> functions regcomp() and and regsub() call regerror(), which isn't
> defined anywhere. As a quick fix, I put a line like
>
> #define regerror(p) fprintf(stderr, p)
makefile, which asks you what regexp version you are using. If I remember
right, the MINIX patches are only needed due to a bug in the C compiler.
If you want to complete you library, check lib/portable on the same
server. I wonder that you still didn't need vfprintf, it was the first
thing I discovered after upgrading to 1.5 a year ago.
Michael
Bruce Evans
>
>I managed to compile "less" last night. I used the version on
>_regerror was an undefined symbol. It turns out the the library
>defined anywhere. As a quick fix, I put a line like
>
> #define regerror(p) fprintf(stderr, p)
>Am I missing some part of the library? I'm sure that regerror() is
>meant to be defined somewhere, but I never saw it anywhere. Is this
I think regerror is supposed to be defined by the application. On my
disk there are 3 regerror.c's, for cawf, egrep (on old version by Woods)
and less (hmmm, I think it came with the original sources - version 97).
These all use a simple version with an fprintf like the above (it
should use "%s\n", p).
You can put a dummy version in the library to fix incomplete applications.
--
Bruce Evans (b...@runxtsa.runx.oz.au)