Gpc

gpc@gnu.de

5400 discussions

GPC reference
by Fred 09 Aug '97

09 Aug '97

I'm new to GPC, but have done a bit of programming with Pascal over the last 12 years with Turbo Pascal (3.x to 5.5) and VAX-Pascal on VMS. I'd like to find something approximating a complete reference and would greatly appreciate any help in this quest. I'm more interested in reference VS instruction material at this time. Fred pismo(a)mclv.net

1 0

Re: Objects Bug
by Peter Gerwinski 09 Aug '97

09 Aug '97

Hello, again! According to Bernhard Tschirren: > > After hours of frigging around with HUGE test programs, I accidentally > stumbled on a stupidly simple test program! Even better - it crashes the > compiler! I only hope that you can reproduce the bug. I just fixed it. :-) The problem was that methods get an implicit `Var' parameter called `Self', so they need a pointer to the object be defined. Some centuries ago, I told GPC to define `ptr_to_Myobject = ^MyObject' implicitly (with the first casing being as written, i.e. inaccessible for the user program), so this did work in a program. What I had forgotten was to *export* `ptr_to_Myobject' if this was done in a Unit or Module. Diff follows. A new GPC beta will follow soon. (Some things in the run time library have to stabilize first.) For the impatient: I am uploading the current "state of the art" as a sequence of gzipped diffs to ftp://agnes.dida.physik.uni-essen.de/gnu-pascal/beta/ (without further comment). Thanks for the nice test program, Peter Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer peter.gerwinski(a)uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201] maintainer GNU Pascal [970714] - http://home.pages.de/~gnu-pascal/ [970125] 8< --------------------------------------------------------------------------- --- ../../gpc-970806/p/util.c Wed Aug 6 14:47:15 1997 +++ util.c Sat Aug 9 13:39:50 1997 @@ -1526,8 +1526,7 @@ void check_object_pointer () { - tree p, ptr = NULL_TREE, d, ptr_name_node; - char ptr_name [256]; + tree p, ptr = NULL_TREE; /* In most practical cases the user program will have * explicitely specified a pointer type pointing to the @@ -1549,6 +1548,9 @@ */ if (!ptr && current_type_name) { + tree d, ptr_name_node; + char *ptr_name = xmalloc (IDENTIFIER_LENGTH (current_type_name) + + strlen ("ptr_to_") + 8); sprintf (ptr_name, "ptr_to_%s", IDENTIFIER_POINTER (current_type_name)); ptr = build_pascal_pointer_type (current_type_name); ptr_name_node = get_identifier (ptr_name); @@ -1558,6 +1560,9 @@ finish_decl (d, NULL_TREE, NULL_TREE); current_type_list = chainon (current_type_list, build_tree_list (NULL_TREE, ptr_name_node )); + if (this_is_an_interface_module) + handle_autoexport (ptr_name_node); + free (ptr_name); } }

1 0

Mathematical functions
by Peter Gerwinski 09 Aug '97

09 Aug '97

Hello! The GPC run time library contains a function `_p_sin' which does nothing else than to call the `sin' function from the C library. As a try, I told GPC to forget about `_p_sin' and to call `sin' directly. Here are the results for a test program (compiled with -O3 on Linux): Calling `_p_sin': 9.93user 0.01system 0:10.11elapsed 98%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (72major+14minor)pagefaults 0swaps Calling `sin' directly: 9.11user 0.01system 0:09.12elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (72major+14minor)pagefaults 0swaps This means a gain of speed of about 9%. If there is no special reason for having an extra `_p_sin' I would vote for dropping it in favour of calling `sin' directly, and similar for other mathematical functions. Comments? Yours, Peter Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer peter.gerwinski(a)uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201] maintainer GNU Pascal [970714] - http://home.pages.de/~gnu-pascal/ [970125]

3 2

-g Works
by Bernhard Tschirren 09 Aug '97

09 Aug '97

Compiling with '-g' seems to work. I dont usually compile with this option, so I cant be sure. However, the last couple of programs compiled OK with '-g'. Great Work! Au Revoir ________________________ | _) | | e-mail: bernie(a)icenet.com.au |___)ernhard |_|schirren ptschirrenb(a)cc.curtin.edu.au www: www.geocities.com/CollegePark/Quad/8070/

2 1

Re: Mathematical functions
by Peter Gerwinski 09 Aug '97

09 Aug '97

According to Jan-Jaap van der Heijden: > > `sin()' is not doing any checking, so you could replace it > right away. But there are other functions that do some checking before > calling the underlying OS function (like `ln') > > GCC knows an optimzation switch -ffast-math, [...] What about this: For gpc-2.1: Drop `sin()' (without any alias) and other non-checking functions, but keep the `_p_foo()' variants for functions that *do* some checking. For gpc-2.2: Drop the `_p_foo()' versions of all mathematical functions but those not supported by the OS (read: by `libc.a') and do the checking inline. Make `--fast-math' turn out the checking. (Things like this can be faced together with range checking in general.) Peter Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer peter.gerwinski(a)uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201] maintainer GNU Pascal [970714] - http://home.pages.de/~gnu-pascal/ [970125]

1 0

Re: Objects Bug
by Peter Gerwinski 09 Aug '97

09 Aug '97

Hallo, Bernhard! Hello, everybody! I have found a workaround: According to Bernhard Tschirren: > Type PGlobalObject = ^tGlobalObject; > TGlobalObject = Object > Constructor Init; > Procedure Method1; Virtual; > Procedure Method2; Virtual; > Procedure Method3; Virtual; > End; It seems that GPC has problems with the implicit `Self' parameter of methods if there is no explicit pointer declared for the object. Later, Peter Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer peter.gerwinski(a)uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201] maintainer GNU Pascal [970714] - http://home.pages.de/~gnu-pascal/ [970125]

1 0

Re: {$L} Good - but not good enough!
by Peter Gerwinski 09 Aug '97

09 Aug '97

According to Bernhard Tschirren: > {$L filename} does not work with library files. For example {$L libgrx.a} > does not work because GPC tries to link 'libgrx.o' !? Ups - sorry, I should not produce such trivial errors. (-# Probably I was tired when I wrote that.) 8< -------------------------------------------------------------------------- --- gpc-970806/p/gpc-lex.c Wed Jun 25 01:56:57 1997 +++ gpc/p/gpc-lex.c Sat Aug 9 10:52:26 1997 @@ -232,7 +232,7 @@ while (p > filename && *p != '.') p--; if (*p == '.' - && (p[1] != 'o' || p[1] != 'a' || p[2] != 0)) + && ((p[1] != 'o' && p[1] != 'a') || p[2] != 0)) { /* Extension given, but not a linker input file. * Recompile it with --automake, if necessary. 8< -------------------------------------------------------------------------- One problem remains with this: GPC does not search the `-L' directories, but only checks the current one. Peter Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer peter.gerwinski(a)uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201] maintainer GNU Pascal [970714] - http://home.pages.de/~gnu-pascal/ [970125]

1 0

Re: Size Of Objects - Revisited
by Peter Gerwinski 09 Aug '97

09 Aug '97

According to Bernhard Tschirren: > You cannot use SizeOf with an object TYPE. See test program below: > [...] Now you can. :-) Diff below. > Sayonara Do swidanya, Peter Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer peter.gerwinski(a)uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201] maintainer GNU Pascal [970714] - http://home.pages.de/~gnu-pascal/ [970125] 8< --------------------------------------------------------------------------- --- ../../gpc-970806/p/parse.y Wed Aug 6 14:47:15 1997 +++ parse.y Sat Aug 9 11:30:51 1997 @@ -5238,14 +5240,30 @@ && DECL_BIT_FIELD (TREE_OPERAND ($3, 1))) error ("`SizeOf' applied to a packed record field"); + /* Now $3 is either a variable access or a type declaration. + * In both cases, TREE_TYPE ($3) carries the actual type. + */ if (PASCAL_TYPE_OBJECT (TREE_TYPE ($3))) { /* Read the size of the object at run time from the VMT. */ - tree vmt = build_component_ref ($3, - get_identifier ("vmt")); - $$ = build_component_ref (build_indirect_ref (vmt, - "`SizeOf'"), + tree vmt; + if (TREE_CODE ($3) == TYPE_DECL) + { + char *name = concat ("", "vmt_", IDENTIFIER_POINTER ( + TYPE_LANG_NAME (TREE_TYPE ($3)))); + vmt = lookup_name (get_identifier (name)); + if (! vmt) + abort (); + free (name); + } + else + { + vmt = build_component_ref ($3, + get_identifier ("vmt")); + vmt = build_indirect_ref (vmt, "`SizeOf'"); + } + $$ = build_component_ref (vmt, get_identifier ("object_size")); } else

1 0

Re: Objects Bug
by Peter Gerwinski 09 Aug '97

09 Aug '97

According to Bernhard Tschirren: > > After hours of frigging around with HUGE test programs, I accidentally > stumbled on a stupidly simple test program! Even better - it crashes the > compiler! I only hope that you can reproduce the bug. I can. :-) Okay, let's try to fix it ... Peter Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer peter.gerwinski(a)uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201] maintainer GNU Pascal [970714] - http://home.pages.de/~gnu-pascal/ [970125]

1 0

Re: Incorrect version display
by Peter Gerwinski 09 Aug '97

09 Aug '97

According to Bernhard Tschirren: > > I have installed GPC-970714 for DJGPP. However, typing 'GPC -v' at the > command prompt shows the following information: > > Reading specs from e:/djgpp/lib\specs > gpc version 970624(2.7.2.1) > > WRONG VERSION! What happened? Probably my fault: I forgot to change this for DJGPP, too, before I compiled gpc-970714. My current DJGPP source tree is correct here, but I don't remember whether I fixed it in time or afterwards. ;-) Peter Dipl.-Phys. Peter Gerwinski, Essen, Germany, free physicist and programmer peter.gerwinski(a)uni-essen.de - http://home.pages.de/~peter.gerwinski/ [970201] maintainer GNU Pascal [970714] - http://home.pages.de/~gnu-pascal/ [970125]

1 0

Jump to page:

2025

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