[fpc-devel]Re: Bug in AssignPipe() on Linux

[Marco van de Voort]

> > It is a FPC/FCL/LCL/Lazarus bug, but mostly fpc.
> 
> I don't use Lazarus  - ironically, I ran into this bug while
> trying to recompile a program that interfaces FPC with NEdit.
> It doesn't use Lazarus or GTK, but it uses popen(F, 'ppc386 -iV', 'R') 
> to check the compiler version. As a temporary workaround, I can use 
> oldlinux.popen() or shell('ppc386 -iV > version.txt') and read the 
> result from there.
> 
> Thanks a bunch for your help - I would never have guessed
> the timezone file had anything to do with opening a pipe!

The real, more fundamental problem a convention change about 
errorhandling. 

A side effect is that uncatched errors internally in the RTL can propagate
to the next call. 

The actual (convention) problem was that assignpipe read "errno" when the
call hadn't failed. This was no problem in the old convention
(errno/linuxerror reset to zero on succesfull call) but is in the new one.
  
> PS: For some reason my posts never get nested in the right 
> thread - Maybe pipermail doesn't like yahoo's headers ?

Maybe the yahoo client or server mutilates some headers. I don't use
threading in email atm, so I wouldn't know.