Gpc June 1999

gpc@gnu.de

22 participants
50 discussions

Problemn with gpc
by root 04 Jun '99

04 Jun '99

I ported some code to gpc 2.0, i make the test in home with a linux rh 5.2, whit this piece of code, all works fine: program test; procedure test( s: string); begin writeln('String:',s); end; begin test('Esto es de prueba'); end. But when i put the procedure prueba in a unit, it crash: <<modulo.p>> unit modulo; interface procedure prueba( s: string); implementation procedure prueba( s: string); begin writeln('Cadena:',s); end; end. <<prueba.p>> program Prueba; uses modulo; begin prueba('Esto es de prueba'); end. And this is the output: Cadena:E Segmentation fault (core dumped) And this with xxgdb: (xxgdb) list modulo.p:1 (xxgdb) tbreak 11 Breakpoint 1 at 0x8048d93: file modulo.p, line 11. (xxgdb) run Prueba (S={Capacity = 0, length = 17, string = 0xbffffd88}) at modulo.p:11 The diference is te Capacity = 0, with the first code its run fine??? What can i do???? I need this work in gpc 2.0 because is the compiler that have the machine, in that i must drop the practics

2 1

Re: Another runtime question
by Frank Heckenbach 04 Jun '99

04 Jun '99

Jo Dillon wrote: > More runtime porting issues - I notice that the macro WRITE_VARIOUS_TYPES > in write.c has lots of cases where if FIX_WIDTH_MASK is true an extra > argument is popped off the arglist (presumably the width), otherwise > it's ignored. When is FIX_WIDTH_MASK set? Is it on by default for stdout? It's set (by the code generated by the compiler in rts.c: rts_write()) when a width is specified in the Write[ln] statement, i.e. `Write (foo : 42)'. In this case, the width (42) is pushed automatically by the compiler generated code. It's got nothing to do with stdout vs. other files. However, a little warning: as I said in my last posting about `Re: Writeln(real)', we plan to change the calling mechanism substantially (though not in the nearest future, but sometime). When that happens, you'll have to make according changes in your runtime. That's why it seems preferable to keep as much as possible of the RTS and to only replace the low-level functions (e.g. write()) as far as needed... Frank -- Frank Heckenbach, frank(a)fjf.gnu.de, http://fjf.gnu.de/ GPC To-Do list, latest features, fixed bugs: http://agnes.dida.physik.uni-essen.de/~gnu-pascal/todo.html

1 0

Another runtime question
by Jo Dillon 04 Jun '99

04 Jun '99

More runtime porting issues - I notice that the macro WRITE_VARIOUS_TYPES in write.c has lots of cases where if FIX_WIDTH_MASK is true an extra argument is popped off the arglist (presumably the width), otherwise it's ignored. When is FIX_WIDTH_MASK set? Is it on by default for stdout? -- Jo

1 0

Re: Writeln(real)
by Frank Heckenbach 04 Jun '99

04 Jun '99

Oliver Puetz wrote: > outputs > > 0.000 0.000 > > but it should be > > 10.000 10.000 > > gpc without any parameters reports > > Integers and strings will be written correctly. I've tested other sizes of > floating-point-vars, but nothing helps. That's the GCC (and EGCS) bug under IRIX that Miklos mentioned. It's a bug with real numbers as varargs which breaks Write[ln] of real numbers in GPC. There are plans to do Write[ln] without varargs in GPC, but since there are many other things to do, don't hold your breath... So, the better thing to do is to report the problem to the EGCS maintainers and let them fix the bug. Here's a demo C program that shows this bug (writes 0 (or perhaps some random number) instead of 8 under IRIX) which you can send them to report the bug. #include <varargs.h> #include <stdio.h> void foo (va_alist) va_dcl { va_list p; double q; va_start (p); q=va_arg (p, double); printf ("%i\n", (int) q); } int main () { double q = 8.0; foo (q); return 0; } Frank -- Frank Heckenbach, frank(a)fjf.gnu.de, http://fjf.gnu.de/ GPC To-Do list, latest features, fixed bugs: http://agnes.dida.physik.uni-essen.de/~gnu-pascal/todo.html

1 0

Writeln(real)
by Oliver Puetz 04 Jun '99

04 Jun '99

Hi, following program Test Program Test; var a :double; begin a:=10.0; Writeln(a :10:3, 10.0 :10:3); end. outputs 0.000 0.000 but it should be 10.000 10.000 gpc without any parameters reports ld32: WARNING 84: /usr/lib32/libm.so is not used for resolving any symbol. ld32: WARNING 85: definition of opterr in /tmp/puetzobq/usr/local/lib/gcc-lib/mi ps-sgi-irix6.2/2.8.1/libgpc.a(getopt.o) preempts that definition in /usr/lib32/m ips3/libc.so. ld32: WARNING 85: definition of optind in /tmp/puetzobq/usr/local/lib/gcc-lib/mi ps-sgi-irix6.2/2.8.1/libgpc.a(getopt.o) preempts that definition in /usr/lib32/m ips3/libc.so. ld32: WARNING 85: definition of optopt in /tmp/puetzobq/usr/local/lib/gcc-lib/mi ps-sgi-irix6.2/2.8.1/libgpc.a(getopt.o) preempts that definition in /usr/lib32/m ips3/libc.so. ld32: WARNING 85: definition of optarg in /tmp/puetzobq/usr/local/lib/gcc-lib/mi ps-sgi-irix6.2/2.8.1/libgpc.a(getopt.o) preempts that definition in /usr/lib32/m ips3/libc.so. Integers and strings will be written correctly. I've tested other sizes of floating-point-vars, but nothing helps. Regards Oliver -- Oliver Puetz Tel.: 0234-700-2982 Lehrstuhl fuer Fluidenergiemaschinen Fax.: 0234-7094-358 Ruhr-Universitaet Bochum, Universitaetsstr. 150, 44801 Bochum opuetz(a)fem.ruhr-uni-bochum.de Gebaeude IB 4/141 www.ruhr-uni-bochum.de/www-public/volgmwbr/mitarb/puetz.htm

1 0

Re: GNU Binary for SGI
by Oliver Puetz 04 Jun '99

04 Jun '99

On Fri, 4 Jun 1999 11:33:15 +0200, Peter Gerwinski wrote: > uses > GPC; (* `MaxLongInt' is defined here *) with the environmental settings, gpc will now be compiled. Thanks! Regards Oliver -- Oliver Puetz Tel.: 0234-700-2982 Lehrstuhl fuer Fluidenergiemaschinen Fax.: 0234-7094-358 Ruhr-Universitaet Bochum, Universitaetsstr. 150, 44801 Bochum opuetz(a)fem.ruhr-uni-bochum.de Gebaeude IB 4/141 www.ruhr-uni-bochum.de/www-public/volgmwbr/mitarb/puetz.htm

1 0

Re: GNU Binary for SGI
by Oliver Puetz 04 Jun '99

04 Jun '99

On Fri, 4 Jun 1999 11:33:15 +0200, Peter Gerwinski wrote: Hi Peter, > Program Test; > > uses > GPC; (* `MaxLongInt' is defined here *) while compiling gpc.pas gpc reports gpc.pas:73: type name expected, identifier `Cstring' given gpc: Internal compiler error: program gpc1 got fatal signal 11 gpc1: gpc exited with status 1 parameters was gpc -B ... -L ... --automake I had to copy gpc.pas into my local dir, otherwise it wouldn't be found. adding -I /tmp/puetzobq/usr/local/lib/gcc-lib/mips-sgi-irix6.2/2.8.1/units/ as gpc.pas its directory didn't help. Could it be that all errors occur, because setting the environmental variables like GPC_EXEC_PREFIX didn't seem to work? Regards Oliver -- Oliver Puetz Tel.: 0234-700-2982 Lehrstuhl fuer Fluidenergiemaschinen Fax.: 0234-7094-358 Ruhr-Universitaet Bochum, Universitaetsstr. 150, 44801 Bochum opuetz(a)fem.ruhr-uni-bochum.de Gebaeude IB 4/141 www.ruhr-uni-bochum.de/www-public/volgmwbr/mitarb/puetz.htm

2 1

Re: GNU Binary for SGI
by Klaus Espenlaub 04 Jun '99

04 Jun '99

Oliver Puetz wrote: > It seems not to work. I've made the settings like you wrote in my .cshrc > setenv reports this: > > GPC_EXEC_PREFIX=/net/sgi249/share/mips/gnu/lib/gcc-lib/mips-sgi-irix6.2/2.8.1/ > CC_LIBRARY_PATH=/tmp/puetzobq/usr/local/lib/gcc-lib/mips-sgi-irix6.2/2.8.1/ [somehow the first character of the two variable names got lost. I assume they were there...] Reading the docs and doing a "strings gpc" it turns out that the env variable name is wrong: it's still GCC_EXEC_PREFIX. Maybe this should be fixed or documented. And the libraries are searched at LIBRARY_PATH. The value for LIBRARY_PATH looks correct for me, but GCC_EXEC_PREFIX should be set to net/sgi249/share/mips/gnu/lib/gcc-lib/ in your case. Try gpc -print-search-dirs to check if the settings are now accepted. And don't forget that these settings will affect the GNU C compiler as well! I'd expect gcc to break if these settings are in place. One last word about LIBRARY_PATH: I'd use -L if I am planning to use a hacked version of the library, as it seems the case in your settings. Klaus

2 1

GNU Binary for SGI
by Oliver Puetz 04 Jun '99

04 Jun '99

Hi Peter, >Sanity check: Did you export or perhaps only set the environment >variable? With `bash', the difference is > > GPC_EXEC_PREFIX=/gpc/usr/local/lib/gcc-lib/mips-sgi-irix6.2/2.8.1 # wrong > >vs. > export GPC_EXEC_PREFIX=/gpc/usr/local/lib/gcc-lib/mips-sgi-irix6.2/2.8.1 # correct First I've set them only via ...=..., but under a cshell, where no export works. with bash, I can export them (export shows the correct values), but only with gpc -B ... -L ... test.pas it runs now :-))) Thank you very much I've got some problems with porting programs from borland to gpc. E.g. in the online-manual http://agnes.dida.physik.uni-essen.de/~gnu-pascal/gpc_5.html#SEC71 is listed, that with gpc --borland-pascal maxint and maxlongint will be predefined. But they aren't. What's wrong? Thanks in advance Oliver -- Oliver Puetz Tel.: 0234-700-2982 Lehrstuhl fuer Fluidenergiemaschinen Fax.: 0234-7094-358 Ruhr-Universitaet Bochum, Universitaetsstr. 150, 44801 Bochum opuetz(a)fem.ruhr-uni-bochum.de Gebaeude IB 4/141 www.ruhr-uni-bochum.de/www-public/volgmwbr/mitarb/puetz.htm

3 2

Re: GNU Binary for SGI
by Klaus Espenlaub 04 Jun '99

04 Jun '99

Oliver Puetz wrote: > > export GPC_EXEC_PREFIX=/gpc/usr/local/lib/gcc-lib/mips-sgi-irix6.2/2.8.1 > # correct > > First I've set them only via ...=..., but under a cshell, where no export > works. with bash, I can export them (export shows the correct values), but > only with > > gpc -B ... -L ... test.pas > > it runs now :-))) Thank you very much if you're using csh then the setting of an environment variable (csh uses a different construct for local shell variables) looks like setenv GPC_EXEC_PREFIX /gpc/usr/local/lib/gcc-lib/mips-sgi-irix6.2/2.8.1 Important: no "=" between the env variable name and its value. Klaus

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

Gpc June 1999