This is the mail archive of the cygwin-developers mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: FW: Random fork failures


On Aug 25 08:22, Carter, Mark Andrew (Andy) wrote:
> We are experiencing an increasing frequency of fork failures in moving
> our 32 bit application to Windows 10.  Rebaseall is only a temporary
> fix.  Preliminary unit testing using this
> "https://github.com/s-u/multicore/blob/master/src/forknt.c"; code have
> been successful.  Would it be fruitful to replace the Cygwin fork with
> this implementation?  I do not want to waste my time repeating
> another's work or pursuing a known dead end.

This implementation won't work as desired.  It's a nice idea and I tried
and tested something similar once.  The downside is that, even if the NT
kernel can fork, the Win32 subsystem doesn't allow this to be done
cleanly.  For instance, after a fork, the connection to the current
Windows console, user32 and gdi32 initialization are broken in the child
and more along these lines.  I talked to guys at Microsoft about this
and they basically waved their hands.

Move to 64 bit Cygwin is what you should do.

If that doesn't work for you, reduce the number of installed DLLs to
lower the footprint and make sure to reduce BLODA interference.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

Attachment: signature.asc
Description: PGP signature


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]