I want to warn any developper that release 1.0.2 has a bug in its stabs generation
that make the GNU debugger GDB (or the IDE with integrated debugger crash)

This problem does not exist with version 1.00, I introduced it while trying to solve
a minor problem that local classes were not understood correctly by the debugger.

We are working on a secure fix for this bug but I strongly advice people NOT to use 1.0.2 to debug FPC code containing classes.

   This warning only applies to Classes (Delphi like) not to Objects (BP like).
BP like objects debugging should work without problems in version 1.0.2;
    I apologize deeply for this major (in my opinion at least) bug!
   Anyway, a version 1.0.4 will be needed sooner or later!

(I cannot tell you the exact situation of the snapshots regarding this problem, it might also still contain some problems)

Pierre Muller
Institut Charles Sadron
6,rue Boussingault
F 67083 STRASBOURG CEDEX (France)

Phone : (33)-3-88-41-40-07 Fax : (33)-3-88-41-40-99

Hello

Michael, when you send subj to the cvs?

Sorry, for offtopic (may be?)

Thanks.

Michail A.Baikov wrote

Hello Michael, when you send subj to the cvs?

It is in cvs; Sorry I was very busy yesterday. We had a demo on my work and about everything went wrong :/

Michael.

Everydays on ftp.ru.freepascal.org/fpc/russian/fpc-cygnus/

In this archive i'm placed lastest version of all utilities, which use in fpc.

May be in future developers change old (?but stable) utilities on new version (with new improvements,but might be not stable). All might be ... ;)

List of utilities and them version (on today):

cigwin1.dll - v1.1.4 (fpc use ??)
arw.exe - GNU ar 2.10.90 (fpc use 2.9.5)
asw.exe - GNU as 2.10.90 (fpc use same version)
dlltool.exe - GNU dlltool 2.10.90 (fpc use 2.9.5)
ldw.exe - GNU ld 2.10.90 (fpc use 2.9.5)

and etc ...

fpc use new version of as.exe, but other program is old.

It's IMHO only ;-)

Everydays on ftp.ru.freepascal.org/fpc/russian/fpc-cygnus/ In this archive i'm placed lastest version of all utilities, which use in fpc.

fpc has mingw32 utils that don't require cygwin1.dll. The cygwin1.dll is only used for the win32 ide because the libgdb uses it.

i know, i am say about only win32 (cygnus) version of fpc.

Michail A.Baikov wrote

i know, i am say about only win32 (cygnus) version of fpc.

It seems you miss the point: there is NO cygnus version of fpc;
the mingw32 version is different from the cygwin version.

Mingw32 is a 100% native version of the gnu utils on Windows cygwin uses a 'unix' emulation library in cygwin.dll

We use the mingw32 version, since this doesn't require a DLL.

Michael.

make install

and report:

C:\pp\source\rtl\win32>make install
ginstall -m 755 -d /pp/units/win32/rtl
cp -fp wprt0.ow wdllprt0.ow /pp/units/win32/rtl
C:/PP/BIN/WIN32/cp: /pp/units/win32/rtl/wprt0.ow: Input or output error (EIO)
C:/PP/BIN/WIN32/cp: /pp/units/win32/rtl/wdllprt0.ow: Input or output error (EIO)

make.exe: *** [fpc_loaders_install] Error 1

Might be better verify platform? I am use NOT *NIX, and Windows2000 Advanced Server.

C:\pp\cvs>cvs -d -z3 checkout fpc
cvs server: Updating fpc
cvs server: Updating fpc/api
cvs server: Updating fpc/api/freebsd
cvs server: failed to create lock directory in repository `/FPC/CVS/fpc/api/free
bsd': Permission denied
cvs server: failed to obtain dir lock in repository `/FPC/CVS/fpc/api/freebsd'
cvs [server aborted]: read lock failed - giving up

May be server updating now?

Fixed.

Michael.

Are the file/directory permissions set correctly ? What user are you doing this as Administrator or another user ?

Michael.

Typical error if you use the dos assembler to create win32 files.

Did you try to patch the file by hand?

In my case (see my email of log), I'm running a dos window under w95... & as normal for the daily snapshots, using my 4dos bat files to make w32, with the make files in fpc.zip, using v1.0.2 go32 as initial compiler... Just looks to me like a w32 compilation error eg related to smartlinking?? ...

I am administrator and all problem in NOT security, it's might be curved (linux-like) makefile... IMHO.

in win32 demo directory (v1.0.2):

E:\pp\source\demo\win32>make all
ppc386.exe !FPCEXTCMD winhello.pp
Note: Reading further options from (env) FPCEXTCMD Free Pascal Compiler version 1.0.2 [2000/10/12] for i386 Copyright (c) 1993-2000 by Florian Klaempfl
winhello.pp(16,10) Hint: Unit STRINGS not used in WINHELLO ppc386.exe !FPCEXTCMD menu.pp
Note: Reading further options from (env) FPCEXTCMD Free Pascal Compiler version 1.0.2 [2000/10/12] for i386 Copyright (c) 1993-2000 by Florian Klaempfl
menu.pp(60,13) Hint: Unit STRINGS not used in MENUDEMO ppc386.exe !FPCEXTCMD dlltest.pp
Note: Reading further options from (env) FPCEXTCMD Free Pascal Compiler version 1.0.2 [2000/10/12] for i386 Copyright (c) 1993-2000 by Florian Klaempfl
dlltest.pp(15,24) Fatal: Syntax error, : expected but const string found make.exe: *** [dlltest.exe] Error 1

BUT! If add -S2 option - all right.

What is -S2 (ObjPas) ? link to docs please ;) and correct please this error.

main program:

Program TestDLLBug;
Uses    Strings,
        Windows;
type
        xreadconfig = function (var s:string;var s1:string):boolean;
var
        pc              :       Pchar;
        ReadConfig      :       XReadConfig;
        Handle          :       THandle;
        S, S1           :       String;
begin
  pc:=StrAlloc (length('C:\CORE2.DLL')+1);
  pc:=StrPCopy(pc,'C:\CORE2.DLL');
  Handle:=LoadLibrary(pc);
  Writeln(Handle);
  if Handle >= 32 then
  begin
   @readconfig:=GetProcAddress(Handle,'READCONFIG');
   if @readconfig=nil then
   begin
    Writeln('CORE2.DLL missed or corrupt! Program halt.');
    halt(252)
   end;
  end;
 writeln('Begin readconfig');
 readconfig(S,S1);
 writeln('End readconfig');
 FreeLibrary(Handle);
end.

CORE2.DLL:

library core2;
uses windows;
function readconfig (var S:String;var S1:String):boolean;export;stdcall;
begin
assign(output,'');
writeln('Hello!');
close(output);
end;
exports
        readconfig;
begin
end.

And this not work.

1. Why not work assign(output,'');writeln('blabla');close(output); ? In BP70/DPMI all working
2. Why in -S2 option fpc don't allow assign values to an address (@readconfig := blablabla (see up)).

Screen after runed testdll.exe:

268435456
Begin readconfig
Runtime error 103 at 0x10001085
0x10001085
0x00401158
0x004012D9

Runtime error 216 at 0x02CDFC06
0x02CDFC06
Runtime error 216 at 0x0040234E
0x0040234E
0x00401CB8
0x004016EE
0x0040152F
0x004023A4
0x02CDFC06

Runtime error 103 at 0x10001085
0x10001085
0x81702954
0xBFFB1B20
0xBFFB1B20

2. Why in -S2 option fpc don't allow assign values to an address (@readconfig := blablabla (see up)).

Because that's ugly TP/Delphi syntax :) Think about it, such a statement doesn't make sense: you assign a value to an address of a variable! (I know in TP it does mean something else). Use

readconfig := @blablabla

in FPC instead (or use -Sd instead of -S2).

Jonas

Michail A.Baikov wrote

I am administrator and all problem in NOT security, it's might be curved (linux-like) makefile... IMHO.

It is possible, but doubtful; I use NT myself and it works fine ?

Michael.

AFAIK you cannot write from a dll to standard output. I will test this with Delphi tomorrow.

2. Why in -S2 option fpc don't allow assign values to an address (@readconfig := blablabla (see up)).

Because -S2 is not 'TP' or 'Delphi' syntax; it is ObjFPC syntax. (And a statement where you assign to an address is strange IMHO)

Michael.

I am understand. FPC don't require open standart input (for readln function) and standart output (for writeln function). All simplest ;)

But why fpc exports from DLL all names in UPPER register? BP, imho don't touch export names. In source small characters, and in exports section small also.

It's not strange. VP/Delphi understand this construction.

It's: "in address of readconfig variable (see source) we take address from GetProcAddress"

Michael, it's "high aerobatics" in Pascal :-) I am too not all understand in this technology (callback and dyno-plugins technology) :-)

I'm don't know, but my Windows stay standalone in the network.Without domain, without users (and local users). Only me :-) This month i'm everyday want change AdvServ to simple Proffesional :) but .... time is over ... :-/

What is?

ftp://www.ru.freepascal.org :-(

Please Daniel, correct this stupid errors:

in sdown.html -> ftp://www.ru.freepascal.org to http://www.ru.freepascal.org

and in links.html -> ftp://www.ru.freepascal.org to ftp://ftp.ru.freepascal.org (if domain exist)

Big Thanks!

On Wed, 18 Oct 2000, Michail A.Baikov wrote:

Zachem nam blini ?

Anyway, it's fixed, please check on www.freepascal.org.

Michael.

I am know, what you know russian language ;-)

Thanks for help.

Hello to all,

I've just seen the new functions in dos Ide for folding lines.

As it is in earlier development stage, I think this is the best moment to express my opinions. By me, I would prefer a more automated way to fold lines, ie. move the cursor at the beginning of a line, and press a key to fold the following lines until a match condition is encountered. If the first line contains a Procedure or Function keyword, then a matching END is searched for. If the line contains a comment char, like (*, then fold until the closing one. The same, may be, could be applicable to other pascal blocks (records and objects).

Moreover, it would be very interesting to SAVE the file mantaining the current folds. May be using a special sequence in comments, for example (*$FOLD*)...(*$FOLDEND*). This way, it could possible to include a lot of comments in a source, without having to scroll them out when editing. In case we had the possibility to save, an easier way to select the lines to fold becomes less important - for every file the work would be done just once.

Another nice thing would be nested folds. This way, for example, an entire INTERFACE section could be folded, still containing subfolders for types and other declarations.

Another last thing (honest): how about let the user define keystrokes? I mean, I hate to type Ctrl-Q-F to search for text (I prefer Ctrl-F). This is not difficult to do like the points before.

Let me know about my ideas - thanks and cheers.

__________________________________________________ Do You Yahoo!?
Talk to your friends online with Yahoo! Messenger. http://im.yahoo.com

At 14:07 19/10/00 , vous avez écrit:

If its not difficult, why don't you code it your self please! We would be very happy to get more developpers for the IDE !

Let me know about my ideas - thanks and cheers.

Ideas are good, developping take much more time than just giving ideas...
But I indeed agree that configurable shortcut keys are a good idea!

Pierre Muller ha scritto:

Hey, 3 things:

1) I did'nt say "it is not difficult", I said "not difficult like the other points". This is different.
2) The whole message was not intended as a bug report, so don't you get angry. My message was suggestions and ideas (clearly stated at the beginning). I did not even mention the fact that folding functionality is unuseable because a bug (I think I am not the only one who had seen this error).
3) If this is your own way to put things, ok. I understand.

I am a programmer like you, and this means two things: first, I have a job that takes away almost all of my time. Second, I know what I (and you) feel when someone criticizes your work. We know very well.

For you to know, I would be very happy to help. I expressed this even before in earlier messages, if you remember. Simply I don't have time right now - may be the next year. There is more: I have read "the Ide cannot be compiled, because it uses code not publicly available" (may be turbovision or whatever it calls). If something has changed, let me know. If the Ide was a simpler project, may be I could try, but I think I would loose more time just to understand than to write something useful. The temptation is there... I downloaded the Ide sources - if you tell me that I simply have to compile (i.e. ppc386 fp.pas), I will put configurable keys in it. This is my bet.

Have a nice day, and long life to freepascal!

Actually, we're planning to (optionally) generate the folds structure directly from the parser tree. However, this requires major modifications in the compiler. These modifications are currently in progress, but will take some until the IDE can make use of them.

I don't think that storing fold information in the source code itself would be a good idea. Storing fold information is obviously a trivial improvement, but I think we should store it either in a separate file (for ex. in test.fld for test.pas) or in a "common storage area" (like the desktop file).

Another nice thing would be nested folds. [...]

Nested folds can be created already.

Another last thing (honest): how about let the user define keystrokes? I mean, I hate to type Ctrl-Q-F to search for text (I prefer Ctrl-F). This is not difficult to do like the points before.

This is a good idea, but isn't as easy to solve as you may think. I've put it on my to-do list.

Gabor

Sting wrote

Actually, we're planning to (optionally) generate the folds structure directly from the parser tree. However, this requires major modifications in

How does it mean this? I'm not sure to understand, but if you are saying something like correlate folding with pascal syntax, I'm not sure this is a good idea...

But this arises the problem of consistency between separate files. The other files (*.fld) will probably contain line numbers from the source, and what would happen if the source file is changed by other means?

Great. But precisely, what is the problem about compiling the Ide? I offered myself to help in some way. It is possible to include a precompiled version of that parte of code that is not free? Or any other way to let compiling possible?

Regards.