February 2004 - Gpc

Re: internal error in gpc
by Adriaan van Os 02 Apr '04

02 Apr '04

In the Macintosh Pascal mailing list (<http://pascal-central.com/maillist.html>) Scott Lindhurst wrote: > I just started playing around with gpc, working on getting some old > code to work with it, and have found a bug. This is version 3.3.2d1, > just downloaded a few days ago with the accompanying GPCInterfacesB2, > on OS X 10.2.8. > > I get the message "gpc: Internal error: Illegal instruction (program > gpc1)" when compiling the following program: > > program bug; > > uses > GPCMacOSAll, GPCStringsAll; {No error if remove GPCStringsAll} > > var > tix: UInt32; {No error if change to integer or longint} > begin > tix := 1; > > if tix > TickCount then {error while processing this line} > tix := 2; > end. > > My makefile generates this command line (which is going to be wrapped, > sorry): > gpc --automake --unit-path=./:/Developer/Pascal/GPCPInterfaces/ > --object-path=./build/ --unit-destination-path=./build/ > --executable-path=./build/ > -Wl,-framework,Carbon,-framework,ApplicationServices -o ./bug -Wl,-x > -O2 -Wall -Wno-identifier-case --ignore-garbage-after-dot > --longjmp-all-nonlocal-labels ./bug.pas > > but the internal error can be reproduced with just > gpc --unit-path=/Developer/Pascal/GPCPInterfaces/ ./bug.pas > > I have the workaround of using integer or longint instead of UInt32. Peter N. Lewis helped out by noting that "limit stacksize 8096" solves the problem. This is true, but still it is strange that the compiler needs such a huge amount of stack for compilation. So, I wonder if maybe there is indeed a problem in the compiler, solved only "by accident" by a stacklimit of nearly a megabyte. When I set "limit stacksize 4096", the backtrace reads: Date/Time: 2003-11-07 11:36:46 +0100 OS Version: 10.2.8 (Build 6R73) Host: G4.local. Command: gpc1 PID: 437 Exception: EXC_BAD_ACCESS (0x0001) Codes: KERN_INVALID_ADDRESS (0x0001) at 0xbff7fff0 Thread 0 Crashed: #0 0x000a9dc8 in tree_size (tree.c:161) #1 0x000aa024 in make_node (tree.c:237) #2 0x000c6980 in alloc_node (stringpool.c:70) #3 0x0024b4b0 in ht_lookup (hashtable.c:162) #4 0x000c6b1c in get_identifier (stringpool.c:109) #5 0x0001bb04 in get_operator (expressions.c:425) #6 0x0001c7e4 in build_pascal_binary_op (expressions.c:661) #7 0x0001c514 in parser_build_binary_op (expressions.c:583) #8 0x0000dd48 in yyuserAction (parse.c:4580) #9 0x000136c8 in yydoAction (parse.c:8238) #10 0x00013190 in yyglrReduce (parse.c:8290) #11 0x000124d0 in main_yyparse (parse.c:8991) #12 0x000a49c0 in compile_file (toplev.c:2134) #13 0x000a9c38 in do_compile (toplev.c:5402) #14 0x000a9d04 in toplev_main (toplev.c:5432) #15 0x00001fa8 in _start (crt.c:267) #16 0x00001e28 in start PPC Thread State: srr0: 0x000a9dc8 srr1: 0x0200f930 vrsave: 0x00000000 xer: 0x20000000 lr: 0x000a9dc0 ctr: 0x000c6968 mq: 0x00000000 r0: 0x000aa024 r1: 0xbff80040 r2: 0x6c616773 r3: 0xbff80080 r4: 0x0000002a r5: 0x0000002a r6: 0x00000001 r7: 0x00000002 r8: 0x727900e4 r9: 0x002e8a28 r10: 0x02a4f000 r11: 0x00020000 r12: 0x48042447 r13: 0x00000000 r14: 0x00000000 r15: 0x00000000 r16: 0x00000000 r17: 0x00000000 r18: 0x00387e40 r19: 0x002eb814 r20: 0x00000000 r21: 0x00000001 r22: 0xbff80200 r23: 0x0000002a r24: 0x0000b04b r25: 0xe972469b r26: 0x00883890 r27: 0x00000078 r28: 0x00000000 r29: 0x00011a6c r30: 0x00000001 r31: 0x000a9dc0 [G4:~] adriaan% gpc -v Reading specs from /Developer/Pascal/gpc332d1/lib/gcc-lib/powerpc-apple-darwin/3.3.2/specs Configured with: ../gpc-332d1/configure --enable-languages=pascal,c --prefix=/Developer/Pascal/gpc332d1 --enable-threads=posix --target=powerpc-apple-darwin Thread model: posix gpc version 20030830, based on gcc-3.3.2 The same problem also occurs with the previous alpha and gcc-3.3, but the backtrace is different: Date/Time: 2003-11-07 11:34:03 +0100 OS Version: 10.2.8 (Build 6R73) Host: G4.local. Command: gpc1 PID: 404 Exception: EXC_BAD_ACCESS (0x0001) Codes: KERN_INVALID_ADDRESS (0x0001) at 0xbff7ffe0 Thread 0 Crashed: #0 0x000c3b74 in alloc_page (ggc-page.c:660) #1 0x000c2b24 in ggc_alloc (ggc-page.c:1026) #2 0x0009c430 in make_node (tree.c:292) #3 0x000c25ac in alloc_node (stringpool.c:70) #4 0x0023cd30 in ht_lookup (hashtable.c:161) #5 0x000c2748 in get_identifier (stringpool.c:109) #6 0x000097d4 in get_operator (expressions.c:426) #7 0x0000a478 in build_pascal_binary_op (expressions.c:656) #8 0x0000a1ac in parser_build_binary_op (expressions.c:578) #9 0x0002d8d8 in main_yyparse (parse.c:2400) #10 0x00096e88 in compile_file (toplev.c:2134) #11 0x0009c03c in do_compile (toplev.c:5370) #12 0x0009c108 in toplev_main (toplev.c:5400) #13 0x000018a8 in _start (crt.c:267) #14 0x00001728 in start PPC Thread State: srr0: 0x000c3b74 srr1: 0x0000f930 vrsave: 0x00000000 xer: 0x00000000 lr: 0x000c3b6c ctr: 0x0009c3a0 mq: 0x00000000 r0: 0x000c2b24 r1: 0xbff80060 r2: 0x6c616773 r3: 0x00000023 r4: 0xbff80280 r5: 0x00000027 r6: 0x00000000 r7: 0x00000003 r8: 0xbff80100 r9: 0x002db53c r10: 0x00000001 r11: 0x0000002c r12: 0x88048444 r13: 0x00000003 r14: 0x000000c8 r15: 0xbfffec80 r16: 0x00000034 r17: 0x00000000 r18: 0x0036c9a0 r19: 0x002d94e4 r20: 0x00000000 r21: 0x02fa3e70 r22: 0x00000001 r23: 0xbff80280 r24: 0x00302ad0 r25: 0x00302ad0 r26: 0x00824890 r27: 0x0000008c r28: 0x03423e70 r29: 0x0000002c r30: 0x002fc69c r31: 0x000c3b6c The needed units can be found at <www.microbizz.nl/GPCMacOSAll.pas.bz2> and <www.microbizz.nl/GPCStringsAll.pas.bz2>. They were too large to include with this message. Regards, Adriaan van Os

4 6

interface and implementation
by mkhaldi＠gosympatico.ca 02 Apr '04

02 Apr '04

I have a question about implementation and interface. If i want only to get the interface from a module which has an implementation module, how must I do to get also the symbol init_XXXXX?? example : ---------------------------------------------- module a interface; export a =all; type string80 = string(80); procedure p; end. module a implementation; procedure p; ...... ...... end. module b interface; export b =all; import a; end. module b impleementation; .... end. I just want to get the type definition without the implementation of module a. In the b.o file, init_A symbol appears. ------------------------------------------------- Actually, I compile my modules in 2 steps. The first one is with --interface-only to get gpi file. Then, I compile my module with --implementation-only. Sometimes, I only need the interface files. But, importing a module always adds the init_XXX symbol. So what is the solution?

3 2

packed records' address again
by Francesco Bonomi 31 Mar '04

31 Mar '04

I have been following the "cannot take address of packed record" thread and now I have a similar problem. I perfectly understand that doing a strcopy is an implicit conversion, but I think that the program gets carried away during checks :-) I had a VERY complex data structure and it took me quite a while to reduce it to the minimum, please bear with me if it's anyway a bit intricate. let's have the following definitions: type char_arr = array[1..30] of char; smallrec = RECORD arr : char_arr; END; largerec = packed RECORD one_rec : smallrec; many_rec : array[1..5] of smallrec; one_array : char_arr; many_arrays : array[1..5] of char_arr; END; VAR r:largerec; c:char_arr; The largerec record is packed, and contains an assortment of structures. A lot of assignments are allowed, like: r.one_array :=c; r.one_rec.arr :=c; r.many_arrays[1] :=c; This single one is not allowed: r.many_rec[1].arr:=c; BTW, assigning a single char is allowed: r.many_rec[1].arr[1]:=#0; So I guess it must be a check that deals with implicits type conversions. but: 1) I am not converting strings, I am assigning a char_arr to another char_arr 2) if the "r.one_rec.arr :=c" assignment is allowed, I cannot see how the "r.many_rec[1].arr:=c;" can be more "dangerous" in addressing mode. I am not trying to be fussy just for the fun of it, but my structure is (as life goes :-) just the single variant that will not work. I am porting an old source code to gpc, while keeping the data file compatibility, so I will explore the unpack-access-pack solution, (hoping that it will not degrade performances too much) BTW (and here I go into newbie mode) how can I check if there is after all any physical difference between a packed and an un-packed record? Is there something like a length() to measure a record's size in memory?? Also, do I understand correctly that the proposed solution: > {$pack-struct, maximum-field-alignment 8} > foo = record > sig : array [ 0..3 ] OF Char; > end; > {$no-pack-struct, maximum-field-alignment 0} would change the structure of a "file of foo" ? Thanks! Francesco Bonomi Francesco Bonomi Si.lab s.r.l. Via Citille 13 50022 Greve in Chianti FI- Italy Tel +39 - 055 8544805 Fax +39 - 055 8547418 http//:www.silab.it

6 8

bug (?) in Time Format
by Toby Ewing 07 Mar '04

07 Mar '04

Hi, all There have been some recent improvements to the Mersenne Twister (the random number generator that is used in the GPC unit), and I've been translating them into Pascal for eventual submission to the programming team. As part of this, I wanted to check the execution time, so I wrote a timer (I thought!) into a test program: program testMT; uses GPC, MTrand; {MTrand is the name of the unit I'm working on} const big = 10000000; TimeFormat = ' %Q'; var t1, t2, i, j : medCard; CurrentTime : TimeStamp; begin SeedRand(1234567890); GetTimeStamp (CurrentTime); WriteStr(FormatTime(CurrentTime, TimeFormat), t1); { line 17 } for i := 1 to big do j := randInt; GetTimeStamp (CurrentTime); WriteStr(FormatTime(CurrentTime, TimeFormat), t2); Writeln('Elapsed time: ', (t2 - t1):1); end. The compiler hangs with the following error: testmt.pas(17): Error: reference expected, value given testmt.pas(17): Error: reference expected, value given testmt.pas(17): Error: tree check: expected class 't', have 'x' (error_mark) in require_complete_type, at p/typecheck.c:63 Please submit a full bug report I'm guessing that it's not about my unit, but rather what I'm trying to do with capturing a string into an int... anyway, here is a bug report. Toby

5 11

About GPC's support of func/proc overloading
by Neil Santos 07 Mar '04

07 Mar '04

How does GPC support function/procedure overloading when it allows the declaration of parameters to be skipped when it comes to the implementation of a function/procedure? I read somewhere that a function/procedure declaration in an interface need not be duplicated for the implementation. UIM, the text referred to Extended Pascal. Say function foo is declared like this in the interface section of a module: function foo(bar : integer) : some_return_type; In the implementation, one could just do: function foo; begin {do interesting stuff here} end; Is it possible to overload foo (i.e., to accept a baz parameter, aside from bar) without resorting to having two code blocks with slightly different parameter declarations? And on a slightly related question: Does GPC support, as an extension to Pascal, optional paramaters the way Ada supposedly does? If so, could someone please point me towards the direction I should go in order to find out how? If not, does anyone else think that this would be a useful extension? Thanks for your time. -- - Neil Santos Freelance Programmer & Web Designer - - http://nsantos.f2o.org/ Free Software advocate - -----BEGIN GEEK CODE BLOCK----- Version: 3.12 GAT d+@ s+:+ a? C++++ UL*++++ P++ L++ E--- W+++(--) N-- !o !K w(++) !O M-- V? PS++ PE-- Y++ PGP+ t* 5-- X- R++ tv+ b++>+++ DI+++ D--- G e>+++++ h r++ y? ------END GEEK CODE BLOCK------ Please avoid sending me attachments created with proprietary software. See http://www.fsf.org/philosophy/no-word-attachments.html

4 14

Re: gpc-20030830 for MSYS released
by Gerrit P. Haase 29 Feb '04

29 Feb '04

Hallo Prof, I'm about to build the Cygwin version right now and as with the previous version I also distribute the MinGW components from mingw.org via the Cygwin mirrors (I don't build the MinGW parts myself). Unfortunately the MinGW GCC maintainer Danny Smith doesn't include GPC in his distribution yet. What about a new release of your MinGW pascal compiler? Gerrit -- =^..^=

2 1

Re: bug (?) in Time Format
by Silvio a Beccara 28 Feb '04

28 Feb '04

Hi, I don't know which precision you need for time, but the following unit does all right for me; it can be modified for going down to millisecond resolution. Silvio a Beccara ------------------------------------------------------------------------------------ unit times; interface function times ( tend: boolean; var tinit: double ): double; implementation uses SysUtils, types; function times ( tend: boolean; var tinit: double ): double; var seconds: double; TS : TTimeStamp; DT : TDateTime; begin if tend = false then begin TS:=DateTimeToTimeStamp(Now); tinit := TimeStampToMSecs ( TS ); end else begin TS:=DateTimeToTimeStamp(Now); seconds := ( TimeStampToMSecs ( TS ) - tinit ) / 1000; result := seconds; end; end; end. ------------------------------------------------------------------------------------ | Hi, all | | There have been some recent improvements to the Mersenne Twister (the | random number generator that is used in the GPC unit), and I've been | translating them into Pascal for eventual submission to the programming | team. As part of this, I wanted to check the execution time, so I wrote a | timer (I thought!) into a test program: | | program testMT; | uses GPC, MTrand; {MTrand is the name of the unit I'm working on} | const | big = 10000000; | TimeFormat = ' %Q'; | var | t1, t2, i, j : medCard; | CurrentTime : TimeStamp; | | begin | SeedRand(1234567890); | GetTimeStamp (CurrentTime); | WriteStr(FormatTime(CurrentTime, TimeFormat), t1); { line 17 } | for i := 1 to big do j := randInt; | GetTimeStamp (CurrentTime); | WriteStr(FormatTime(CurrentTime, TimeFormat), t2); | Writeln('Elapsed time: ', (t2 - t1):1); | end. | | The compiler hangs with the following error: | testmt.pas(17): Error: reference expected, value given | testmt.pas(17): Error: reference expected, value given | testmt.pas(17): Error: tree check: expected class 't', have 'x' | (error_mark) in require_complete_type, at p/typecheck.c:63 | Please submit a full bug report | | I'm guessing that it's not about my unit, but rather what I'm trying to do | with capturing a string into an int... anyway, here is a bug report. | | Toby

3 2

None
by None 24 Feb '04

24 Feb '04

+(CET) Received: from adele.gerwinski.de (adele.gerwinski.de [217.69.76.193]) by matsrv.math.cas.cz (8.12.8/8.12.8) with ESMTP id i1O685RC031413 for <jerabek(a)math.cas.cz>; Tue, 24 Feb 2004 07:08:05 +0100 Received: from localhost ([127.0.0.1] helo=gnu.de) by adele.gerwinski.de with smtp (Exim 3.35 #1 (Debian)) id 1AvVcA-0007lD-00; Tue, 24 Feb 2004 07:00:18 +0100 Received: from mtiwmhc11.worldnet.att.net ([204.127.131.115]) by adele.gerwinski.de with esmtp (Exim 3.35 #1 (Debian)) id 1AvVc6-0007kl-00 for <gpc(a)gnu.de>; Tue, 24 Feb 2004 07:00:14 +0100 Received: from yahoo.com +(25.hartford-02rh16rt-03rh15rt.ct.dial-access.att.net[12.76.130.25]) by worldnet.att.net (mtiwmhc11) with SMTP id <200402240558151110000st6e>; Tue, 24 Feb 2004 05:58:15 +0000 Message-ID: <403ADCF2.6EC260D7(a)yahoo.com> Date: Tue, 24 Feb 2004 00:11:14 -0500 From: CBFalconer <cbfalconer(a)yahoo.com> Reply-To: cbfalconer(a)worldnet.att.net Organization: Ched Research X-Mailer: Mozilla 4.75 [en] (Win98; U) X-Accept-Language: en MIME-Version: 1.0 To: gpc(a)gnu.de Subject: [SPAM] Re: About GPC's support of func/proc overloading References: <403A5EED.3030905(a)sexmagnet.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Precedence: bulk Sender: gpc-owner(a)gnu.de X-Spam-Flag: YES X-Spam-Status: Yes, hits=5.3 required=5.0 tests=FAKE_HELO_YAHOO, FORGED_YAHOO_RCVD,RCVD_FAKE_HELO_DOTCOM,RCVD_IN_NJABL,RCVD_IN_SORBS autolearn=no version=2.63 X-Spam-Report: * 1.6 RCVD_FAKE_HELO_DOTCOM Received contains a faked HELO hostname * 2.6 FAKE_HELO_YAHOO Host HELO did not match rDNS: yahoo.com * 0.9 FORGED_YAHOO_RCVD 'From' yahoo.com does not match 'Received' +headers * 0.1 RCVD_IN_SORBS RBL: SORBS: sender is listed in SORBS * [12.76.130.25 listed in dnsbl.sorbs.net] * 0.1 RCVD_IN_NJABL RBL: Received via a relay in dnsbl.njabl.org * [12.76.130.25 listed in dnsbl.njabl.org] X-Spam-Level: ***** X-Spam-Checker-Version: SpamAssassin 2.63 (2004-01-11) on matsrv.math.cas.cz Status: RO Content-Length: 1397 Lines: 43

1 0

Bug#232028: Support for a common place for additional units? (forwarded from Torsten Knodt)
by Matthias Klose 23 Feb '04

23 Feb '04

This report was submitted to the Debian bug tracking system. Is this something generally useful?

2 1

Re: Dos unit on Unix
by Peter Norton 21 Feb '04

21 Feb '04

Thanks to Prof. Abimbola and Frank for their quick and professional help. You helped me to find the right solution ( even two ;-) Best regards, Peter __________________________________ Do you Yahoo!? Yahoo! Mail SpamGuard - Read only the mail you want. http://antispam.yahoo.com/tools

1 0