Hello all,

I'm having some trouble to get BacktraceStrFunc to find line numbers. This is with fpc 3.0.0 on linux x86_64 (Centos 7 if it matters). If I compile the following example with

#fpc -gl tt.pas

I only get this output:

Started...
Exception: $0000000000455540
Done.

Evidently line info is somehow not present here.
Am I doing smth wrong or maybe lineinfo is not yet implemented for x86_64?

Thank you,
Nikolai
============================================================== program tt; {$mode delphi }

uses SysUtils, Classes;
procedure Test3;
var list: TList;
begin
   list := TList.Create;
   if list[10] = nil then
     writeln('Wow?');
end;
procedure Test2;
begin
   Test3;
end;
procedure Test;
begin
   writeln('Started...');
   try
     Test2;
   except
     writeln('Exception: ', BackTraceStrFunc(ExceptAddr));
   end;
   writeln('Done.');
end;
begin
   Test;
end.

No, but the units that we distribute do not have debug information included. So if the error is in the RTL, then there is no debug information.

If I run your program with a RTL compiled with debug info, I get:

home: >./st
Started...
Exception: $000000000046F571 CHECKINDEX, line 56 of ../objpas/classes/lists.inc Done.

Michael.

01.05.2017 11:21, Michael Van Canneyt:
[...]

No, but the units that we distribute do not have debug information included. So if the error is in the RTL, then there is no debug information.

Ok, right, but then I suppose it should show line number as soon as the example is modified like this:

procedure Test3;
var a, b: integer;
begin
   a := 1; b := 0;
   writeln(a/b);
end;

And I'm still getting just an address anyway...

Thank you,
Nikolai

3.0.x is broken in this regard (stack back trace on x86-64 elf targets), see other threads on the fpc mailing lists regarding this. This is why we discussing a quick as possibile 3.0.4 release.

01.05.2017 11:46, Florian Klämpfl:
[...]

Ah, ok. Couldn't find any mentions of the issue (bad googling or whatever) but anyway I'll probably have to wait for 3.0.4 then.

Thank you,
Nikolai

_______________________________________________ fpc-devel maillist - http://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel

You don't need to wait:

Only 3.0.2 linux for i386 CPU has the problem. 64-bit is OK. And 3.0.0 is OK in this aspect for all platforms.

So you can continue to work.

Michael.

OPs, sorry, then I was wrong. I thought x86-64 elf is affected ...

Michael Van Canneyt wrote

Only 3.0.2 linux for i386 CPU has the problem. 64-bit is OK.

hmm, I don't get the lineinfo on x86-64 (Ubuntu 16.04/Mate, 64-Bit). armel and armhf are affected too.

Regards, Bernd.

01.05.2017 14:35, Bernd Mueller:

Personally, for now I'll stick to 2.6.4 -- it works as expected at least on x86_64 linux here.

Besides, I've found my own mistake: when building an rpm package (with pascal program in this case), by default rpmbuild would choose to automagically strip the resulting binaries too much so this behaviour has to be suppressed (E.g. by __os_install_post %{nil} or whatever)

Thank you,
Nikolai

In our previous episode, Bernd Mueller said:

> Only 3.0.2 linux for i386 CPU has the problem. 64-bit is OK. hmm, I don't get the lineinfo on x86-64 (Ubuntu 16.04/Mate, 64-Bit). armel and armhf are affected too.

Please test with the fixes branches. The series of revisions in the bugreport has been merged, but there are still backtrace related fixes unmerged:

http://www.stack.nl/~marcov/mergelogs26/backtrace.txt

BTW: Could you extend your scripts to include links for the revs to ViewVC? Pattern is: https://svn.freepascal.org/cgi-bin/viewvc.cgi?view=revision&revision=35495

In our previous episode, Florian Kl?mpfl said:

I tried, and it is now totally fubarred ;-)

In our previous episode, Marco van de Voort said:

I guess it is something with the apache https changes also working for svn now.

Redirecting
to
URL
'https://svn.freepascal.org/svn/fpc/branches/fixes_3_0':

which messes with my program.

Just relocate the check out?

For people using git-svn, the section 'General Case' here: https://git.wiki.kernel.org/index.php/GitSvnSwitch should help.

In our previous episode, Florian Kl?mpfl said:

I already fixed it immediately after the last msg, without coding even, since the urls were in an ini file. It was just confusing that I recompiled the app with minor mods after a long time, and it bombed. Then you don't immediately expect a server config change :-)

The viewvc revs are also live. Clicking on them might be a bit hard though (depending on browser) because the javascript fold old out code is also active on that part of the line. JS is not my forte.

http://www.stack.nl/~marcov/mergelogs26/backtrace.html

Nice :)

Mine should probably not be merged, but maybe r33038 and most likely r35886?

I tested yesterdays fixes branch and it worked for me now.

Regards, Bernd.

In our previous episode, Bernd Mueller said:

> Please test with the fixes branches. I tested yesterdays fixes branch and it worked for me now.

Good. Most mass merging now have been done. Some dots-on-i and a rereading of various threads for unmerged requests still have to be done.

[...] 3.0.x is broken in this regard (stack back trace on x86-64 elf targets), see other threads on the fpc mailing lists regarding this. This is why we discussing a quick as possibile 3.0.4 release.

Any news here?

Mattias