Re: pipe chunking vs Windows

On Sun, Jul 29, 2007 at 07:43:34PM -0400, Andrew Dunstan wrote:
> 
> 
> Andreas Pflug wrote:
> >Andrew Dunstan wrote:
> >  
> >>I have no idea why that's done - it goes back to the origins of the
> >>syslogger - probably because someone mistakenly thinks all WIndows
> >>text files have to have CRLF line endings.
> >>
> >>I tried changing that to _O_BINARY, and calling _setmode on both the
> >>pipe before it's duped into stderr and stderr after the dup and both.
> >>Nothing seemed to work.
> >>    
> >AFAIR the flag has to be set again in each child process.
> >
> >  
> 
> OK, this was the bit of info I was missing. I put a call in 
> SubPostmasterMain() like this:
> 
>    _setmode(fileno(stderr), _O_BINARY);
> 
> and one in the syslogger code the same, just after the pipe is dup'ed 
> into stderr.
> 
> After that I saw no more corruption issues.

Seems right - for anything EXEC_BACKEND, if you change a state from the
default it won't be inherited, so you nede to explicitly add it in the
child.

> However, the problem is that in backporting it we'd make a slight 
> behaviour change - the log file just gets LF instead of CRLF line 
> endings. I'm inclined to say that's a better result than living with the 
> bug, though.

Can't we add back the CRLF combo when writing the file out from the
syslogger process?

//Magnus