Gpc January 1998

gpc@gnu.de

23 participants
48 discussions

GPC 971001 problems on redhat linux 4.2
by George Brewster 19 Jan '98

19 Jan '98

I am having trouble with GPC 971001 on my RedHat linux box. It doesn't compile _anything_, but always gives the following error, regardless of what I try to compile: /tmp/cca00834.i:1: numeric constant with no digits /tmp/cca00834.i:1: numeric constant with no digits /tmp/cca00834.i:1: invalid #line /tmp/cca00834.i:2: warning: missing program header /tmp/cca00834.i:2: parse error at null character the output of gpc 'trig.pas -v' is: Reading specs from /usr/lib/gcc-lib/i586-unknown-linux/2.7.2.1/specs gpc version 971001(2.7.2.1) /usr/lib/gcc-lib/i586-unknown-linux/2.7.2.1/gpc-cpp -lang-pascal -v -nocharescape -undef -D__GNUC__=2 -D__GPC__=2 -D__GNUC_MINOR__=7 -D__ELF__ -Dunix -Di386 -Dlinux -D__ELF__ -D__unix__ -D__i386__ -D__linux__ -D__unix -D__i386 -D__linux -Asystem(unix) -Asystem(posix) -Acpu(i386) -Amachine(i386) -D__i486__ trig.pas /tmp/cca00815.i GNU CPP version 971001(2.7.2.1) (i386 Linux/ELF) #include "..." search starts here: #include <...> search starts here: /usr/local/include /usr/i586-unknown-linux/include /usr/lib/gcc-lib/i586-unknown-linux/2.7.2.1/include /usr/include End of search list. /usr/lib/gcc-lib/i586-unknown-linux/2.7.2.1/gpc1 /tmp/cca00815.i -quiet -dumpbase trig.pas -version -famtmpfile=trig.gpc -o /tmp/cca00815.s GNU Pascal version 971001(2.7.2.1) (i386 Linux/ELF) compiled by GNU C version 2.7.2.1. /tmp/cca00815.i:1: numeric constant with no digits /tmp/cca00815.i:1: numeric constant with no digits /tmp/cca00815.i:1: invalid #line /tmp/cca00815.i:2: warning: missing program header /tmp/cca00815.i:2: parse error at null character what else can I say?? I did not compile it myself, but used the binary package available. I use a 486 gpc v2.0 worked, but not with the -g flag (that's a whole different problem) I have libc version 5.3.12 I remember hearing it proposed somewhere that this problem might be a result of an old libc, but I haven't heard this confirmed by anybody. I think this was proposed in the gpc 1997 mailing list archive, but I never heard if it was resolved, and I don't subscribe to the list... Currently, I have switched back to gpc 2.0, which seems to work fine... I appreciate any help, George Brewster

1 0

function pointers
by Gerd Balzuweit 19 Jan '98

19 Jan '98

Hi, I want to write some examples in gnu pascal but I have not enough experience in gnu pascal. I want test how powerful the actual version under Win32 is. So I would use some C - functions wellknown in the system. But I need a special type. A pointer of a function. How to do that? Or is that not implemented yet? Something like that: type Pfunc = function( arglist: PVoid ): DWord; May be anyone can help? -- Gerd Balzuweit iXOS Anwendungs-Software GmbH Leipzig E-mail: mailto:gerd.balzuweit@ixos-leipzig.de www : http://www.ixos-leipzig.de/ Phone : +49 341 48503-0 Fax : +49 341 48503-99

1 0

RE: <Autofrwd>NEWS: gcc-2.8.0 released?
by Berend de Boer 19 Jan '98

19 Jan '98

On Saturday, January 17, 1998 4:00 PM, Peter Gerwinski [SMTP:peter@agnes.dida.physik.uni-essen.de] wrote: > Of course, this plan implies that we must upgrade from gcc-2.7.2.x to > gcc-2.8.0 ASAP. I think it is reasonable to continue support for > gcc-2.7.2.x as long as this works without too much effort. The problem is: I use gpc on FreeBSD machines which are on gcc 2.7.2.x. I don't want to upgrade these machines to a gcc different than used to compile the other sources (just to avoid problems). This means I cannot compile gpc sources unless you still support gcc or supply a binary each time. So try to support gcc 2.7.x as long as possible. Groetjes, Berend.

1 0

Re: Interfacing with C!
by Peter Gerwinski 19 Jan '98

19 Jan '98

According to Bernhard Tschirren: > > Are there any ways of calling procedures written in GPC from GCC. Now, > heres the clinch: The pascal source is object oriented!! That's not a real problem. :-) > Now, I have a pointer to the object (in C). The object contains both data There is an implicit `vmt' field, a pointer, in each object. Apart from that, it is just an ordinary `record' (or `struct' for that matter). > and ONLY virtual methods (That makes things simpler!) So, I could easily > get the address of the function that I want to call. Non-virtual methods are even easier: They are called `Myobject_Mymethod' internally and differ from ordinary procedures only by the additional implicit `Self' parameter. To call a virtual method from C you must dig it out: The `vmt' field in the object points to a structure which contains * the size of the object (a `size_type' variable in C), * its negative (another `size_type' variable), * and pointers to the methods in the same order as they appeared in the declaration of the object. To initialize an object, it is not enough to call the constructor, but you must explicitly assign the correct value to the `vmt' field, i.e. let it point to the correct VMT structure. (GPC does this "inline" when a constructor is called.) When you call a method, don't forget to pass the implicit `self' parameter. > So, the question is: How does GPC handle virtual methods? Looking at > numerous assembly listing I have deduced the following: > > GPC passes a pointer to the parent object Why "parent"??? No - that's a pointer to the object itself! > (@Self) NOT on the stack, but > in ECX. Is that true, and how is it implemented on other platforms? This is just by chance and depends on the optimization level. If you declare the `self' parameter as an ordinary (pointer) parameter, it should work just fine because this is exactly what GPC does internally. Good luck, Peter -- Peter Gerwinski, Essen, Germany, free physicist and programmer Maintainer GNU Pascal - http://home.pages.de/~GNU-Pascal/ - 1 Oct 1997 PGP key fingerprint: AC 6C 94 45 BE 28 A4 96 0E CC E9 12 47 25 82 75 Fight the SPAM! - http://maps.vix.com/

1 0

GNU Pascal for gcc 2.8.0
by John Gotts 19 Jan '98

19 Jan '98

Is there a release of GNU Pascal corresponding to gcc 2.8.0 or can I just use one of the alpha or beta releases? John -- John GOTTS <jgotts(a)engin.umich.edu> http://www-personal.engin.umich.edu/~jgotts

1 0

Re: Compability C++ - GPC
by Peter Gerwinski 18 Jan '98

18 Jan '98

According to Hans Ecke: > > Would it be possible to just translate the header files to units and > simple compile the program? [...] Unfortunately not. I thought about planning this for the future, but this would mean that GPC would have to copy multiple inheritance with all its ugly implications from C++; some discussion in this list convinced me that it will be much better to implement "Interfaces" (like in Java) into GPC. > Do the mechanisms of polymorphie, inheritance etc work between a library > written in C++ and an unit written in GPC? They don't. To work around this, you can write a C-style interface that lives without objects for the C++ library and then make a Pascal Unit use these inter- face functions to implement Pascal objects: C++: class foo // the thing we want to use { void bar ( int baz ); } C: /* This non-OOP interface we must write because OOP information * does not survive transport through an `.o' or an `.a' file. */ void foo_bar ( struct class_foo *self, int baz ) { /* call the object's method */ } Pascal: (* This unit implements a `FooObj' that implements a Pascal *) (* OOP interface for the C++ `foo' class. The implementation *) (* uses a non-OOP function written in C. *) Type FooPtr = ^Foo; FooObj = object; cppClass: Pointer; (* holds a pointer to the C++ instance *) Procedure Bar ( Baz: Integer ); end (* FooObj *); Procedure C_foo_bar ( Self: Pointer; Baz: Integer ); AsmName 'foo_bar'; Procedure FooObj.Bar ( Baz: Integer ); begin (* FooObj.Bar *) C_foo_bar ( cppClass, Baz ); end (* FooObj.Bar *); BTW, what about writing Pascal libraries that provide the same functionality? I am myself working on such a thing (BO5); the EFLIB library (http://ftp.sunet.se/EFLIB/) is another approach (I do not know whether it supports X11), and I am sure that many readers of this list have their own libraries waiting to be ported to GNU Pascal. Greetings, Peter -- Peter Gerwinski, Essen, Germany, free physicist and programmer Maintainer GNU Pascal - http://home.pages.de/~GNU-Pascal/ - 1 Oct 1997 PGP key fingerprint: AC 6C 94 45 BE 28 A4 96 0E CC E9 12 47 25 82 75 Fight the SPAM! - http://maps.vix.com/

1 0

Interfacing with C!
by Bernhard Tschirren 18 Jan '98

18 Jan '98

Are there any ways of calling procedures written in GPC from GCC. Now, heres the clinch: The pascal source is object oriented!! Now, I have a pointer to the object (in C). The object contains both data and ONLY virtual methods (That makes things simpler!) So, I could easily get the address of the function that I want to call. So, the question is: How does GPC handle virtual methods? Looking at numerous assembly listing I have deduced the following: GPC passes a pointer to the parent object (@Self) NOT on the stack, but in ECX. Is that true, and how is it implemented on other platforms? Thanks Bernie ________________________ e-mail: bernie(a)icenet.com.au | _) | | ptschirrenb(a)curtin.edu.au |___)ernhard |_|schirren tschirre(a)cs.curtin.edu.au www: www.geocities.com/CollegePark/Quad/8070/

1 0

Compability C++ - GPC
by Hans Ecke 18 Jan '98

18 Jan '98

Hi all! This message asks about writing bindings for a library which is written in C++. I have to write an X-Windows application. I have never done this, but have read about various widgets. The most exciting commentares I read about were about QT from TrollTech. Its written in C++ (I love objects) and there are also quite a bit examples on the net. Of course I would love to write the app in GPC, too. Now my question : Would it be possible to just translate the header files to units and simple compile the program? I.e. something like $> cat test_app.pas programm test_app; uses qtmain; ...body end. $> gpc test_app.pas -lqt $> test_app Do the mechanisms of polymorphie, inheritance etc work between a library written in C++ and an unit written in GPC? I consider this a topic almost as critic as the writing of bindings for a library written in plain C. Hans

1 0

NEWS: gcc-2.8.0 released?
by Peter Gerwinski 18 Jan '98

18 Jan '98

Hi, everybody! I just found this in gnu.misc.discuss (*not* in gnu.announce/gnu.gcc.announc): > From: Paul Eggert <eggert(a)twinsun.com> > Subject: gcc 2.8.0, libstdc++ 2.8.0 released > Newsgroups: gnu.misc.discuss > > ftp://prep.ai.mit.edu/pub/gnu/gcc-2.8.0.tar.gz > ftp://prep.ai.mit.edu/pub/gnu/libstdc++-2.8.0.tar.gz > > More details and an official announcement will appear soon. > Please send bug reports for GNU C to `bug-gcc(a)prep.ai.mit.edu'. > Please send bug reports for GNU C++ to `bug-g++(a)prep.ai.mit.edu'. Even if this was not the official release of gcc-2.8.0, I suggest to let gpc-2.1 follow as soon as possible even if not all bugs are fixed. My current development GPC passes all PVS conformance tests for ISO-7185 Standard Pascal level 0 and (NEW!) level 1 but two minor ones. My idea is to fix these, make the final beta and release gpc-2.1 even if there are some remaining known bugs. (An incomplete list of those bugs that have prevented me from releasing gpc-2.1 up to now follows below.) Of course, this plan implies that we must upgrade from gcc-2.7.2.x to gcc-2.8.0 ASAP. I think it is reasonable to continue support for gcc-2.7.2.x as long as this works without too much effort. Any comments? Peter -- Peter Gerwinski, Essen, Germany, free physicist and programmer Maintainer GNU Pascal - http://home.pages.de/~GNU-Pascal/ - 1 Oct 1997 PGP key fingerprint: AC 6C 94 45 BE 28 A4 96 0E CC E9 12 47 25 82 75 Fight the SPAM! - http://maps.vix.com/ 8< ---- remaining bugs in current GPC --------------------------------------- The two ISO-7185-related bugs I want to fix before releaseing gpc-2.1: * GPC barfs on type checking with Boolean values in `case' statements. * Nested packed conformant arrays do not yet work in all cases. Major bugs I am planning to postpone (unless, of course, I have an idea how to fix them without too much effort;-)-: * The "AutoMake" mechanism needs a complete rewrite. It's too slow and does not always decide correctly whether something must be recompiled or not. * GPC still fails some (few) ISO conformance tests when compiled with optimization on the DEC Alpha. * `Seek' and `SeekRead' etc. do not yet work in all situations (frank95b.pas). * Comparisons of integers of different size sometimes fail (frank118.pas). * String `Var' parameters accept constant values (frank123.pas). * "Procedure Foo ( x: String )" has in fact a `Var' parameter (frank125.pas). * `New' on pointers to discriminated strings doesn't work (frank128.pas). * There is no `--uses="System"' option for full BP compatibility yet. * The `--field-width' switch does not handle `LongInt's etc. * Inside `New', all methods of an object become visible and may cause name clashes.

3 3

Source code released for PROTOTYPE 5.1;
by Johan Larsson 16 Jan '98

16 Jan '98

Hi Everyone, Though prototype 5.1 is an early working model with an incompatible design, I have decided to release source code together with that release. The source code can be downloaded at: http://ftp.sunet.se/EFLIB/Download/ 1,2 MB kernel source code. You are not adviced to use this release of EFLIB in developing since prototype 6 is entirely redesigner. However, you are encouraged to participate in discussions regarding the design chosen in EFLIB prototype 5.1 toward more knowledge about good and bad object-oriented designs. The 6th prototype is scheduled for April 30, 1998 and will include a cross-platform GUI as well as Linux support. Contact me if you want to join the team or want to help a little! Best Regards, Johan Larsson URL: http://www.acc.umu.se/~jola/ MAIL: jola(a)acc.umu.se Also, REMEMBER to check out EFLIB, the free object-oriented framework for EVERY Pascal programmer, at http://ftp.sunet.se/EFLIB/. Simply SOPHISTICATED!

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 January 1998