develooper Front page | perl.perl5.porters | Postings from December 2003

Re: BUG fork/pipe on windows

Thread Previous | Thread Next
Uri Guttman
December 8, 2003 08:22
Re: BUG fork/pipe on windows
Message ID:
>>>>> "NI" == Nick Ing-Simmons <> writes:

  NI> Uri Guttman <> writes:
  >> no one responded to this post 

  NI> Only the ActiveState crowd _really_ understand the windows port,
  NI> and they don't always watch here.

i expected it was an activestate thing but i don't know how to get their
attention to it.

  >> and i think it is important. 

  NI> Lots of stuff on Windows doesn't work as POSIX.

oh, i know that and i worked hard to get this test going on winblows. i
even acquired a new xp box so i could properly test and debug
stuff. asking people to do it remotely for my code was not very

  NI> Patches (even doc patches that say this doesn't work on windows) welcome.
  NI> Also welcome is a complete script which when run on Windows prints
  NI> "not ok"
  NI> and when run on POSIX prints "ok".

i won't patch it until i know for sure that it is a doc bug vs a
internals bug. maybe the open should work on winblows but wasn't
properly done. also that is not a POSIX feature afaik but a winblows
hack to get around open -| and |- not being supported. this whole
section in perlfork is just winblows related.

  NI> There are several things that could be wrong
  NI>   - psuedo fork juggling fds in the parent/child "thread"
  NI>   - Win32 PIPE semantics not being same
  NI>   - trying to use select() on a pipe (doomed...)

select on pipes does work there in the other direction and with
socketpair. this appears to be a bug in the reopen call in the example.



Uri Guttman  ------  --------
--Perl Consulting, Stem Development, Systems Architecture, Design and Coding-
Search or Offer Perl Jobs  ----------------------------

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About