[656] in Pthreads mailing list archive
Re: compile bug in pthread of 16nov97
daemon@ATHENA.MIT.EDU (Database Owner)
Wed May 14 11:56:43 1997
Date: Wed, 14 May 1997 08:45:58 -0700 (PDT)
From: Database Owner <database@galileo.starnetc.com>
To: "Michael J. Gourlay" <gourlay@ucsub.Colorado.EDU>
Cc: proven@MIT.EDU, pthreads@MIT.EDU
In-Reply-To: <199705132118.PAA23035@ucsub.Colorado.EDU>
Just to let you know,
If you're sure that you're running IRIX 6.2, there is a native version of
pthreads for IRIX 6.2 that is available. It works very well. I've been
using it for the past 12 months.
David Green-Seed
david@green-seed.com
On Tue, 13 May 1997, Michael J. Gourlay wrote:
> Hello,
>
> In the process of trying to compile pthreads 1.60 of 16nov97 on an
> SGI Indigo with an IP22 processor running IRIX 6.2 using the native C
> compiler, I received these errors. Please help.
>
>
>
>
> (1) EVERY SINGLE COMPILE LINE during the make reported the following:
> cfe: Warning 581: include/sys/__signal.h:5: Macro sigmask redefined.
> This is because sigmask is already defined in "sys/signal.h" o my
> machine. The definitions seem to be compatible, maybe.
> Here is how my machine defines this:
> #define sigmask(sig) (1L << ((sig)-1))
> Here is the definition in __signal.h:
> #define sigmask(n) ((unsigned int)1 << (((n) - 1) & (32 - 1)))
>
>
>
>
>
>
> (2) in include/pthread/sleep.h the declaration of machdep_gettimeofday
> uses the word "inline" which is obviously not ANSI C. I omitted the
> word "inline" and this no longer caused problems. However, it seems
> very strange that one would want to include a function multiple times
> in a single library so this is not an optimal solution. It would be
> better to rewrite this either as a macro or as a regular function. My
> guess is that this is an attempt to make use of a non-standard feature
> of the GNU C compiler. There was no sign that the code was supposed
> to be C++ code.
>
>
>
>
> (3) This error is much worse:
>
> cfe: Error: pthreads-1_60_beta6/stdio/vfprintf.c, line 188: Syntax Error
> u_quad_t _uquad;
> ---------^
> There were a cacade of other errors in vfprintf.c after this, but
> they were probably due to the above error. This error seems to be
> due to a mess of precompiler code where u_quad_t is only defined if
> either __GNUC__ or __alpha is defined. In case anybody is wondering,
> IRIX 6.2 does define a type called "long long". I defined u_quad_t
> to be "long long", and the compilation proceeded. BTW there is no
> "long double" on this machine.
>
>
>
>
> (4) in stdio/vfprintf.c, the following line causes the warning "Bad
> identifier after the ifndef":
> #ifndef 0
>
>
>
>
> (5) This error killed my buzz. I have no idea how to fix it. I have
> no idea what it means. I guess this is an assembly problem.
> syscall.S is linked to "machdep/syscall-ip22-irix-5.2.S". PLEASE
> HELP:
>
> cc -I. -Iinclude -I/tmp_mnt/mckenna3/local/src/pthreads-1_60_beta6/include -DPTHREAD_KERNEL -g -c syscall.S -o obj/syscall.o
> input buffer on non-int boundry
> gnumake: *** [obj/syscall.o] Error 1
>
>
>
>
> --
> Michael J. Gourlay Physics Graduate Michael.Gourlay@colorado.edu
> Duane Physics D031, Department of Physics, University of Colorado at Boulder
> "That's me; I've been ionized...." - Buckaroo Banzai
> SMAIL: CU Boulder, Physics Department, Campus Box 390, Boulder, CO 80309, USA
> WORK: Colorado Research Associates. 415-9701 FAX: 415-9702
> WEB: http://ucsub.Colorado.EDU/~gourlay
>
