This is the mail archive of the cygwin@cygwin.com 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]

Re: 1.3.2 Catch-22 in tmpnam/tmpfile with gcc -mno-cygwin


Charles Wilson wrote:
> 
> But programs compiled with -mno-cygwin don't use cygwin1.dll (which contains
> newlib, which contains the stdio functions).  -mno-cygwin programs will use the
> msvcrt.dll runtime -- which has its own implementations of the stdio functions.
> Are you saying that msvcrt's version of tmpnam/tmpfile are broken?  If so,
> that's not really our problem...

I've been madly going through the actual source code in question (a
p-code compiler, so it's fairly large, and it's not mine), and begin to
suspect that it's actually bypassing tmpfile (on the grounds that some
supposedly ANSI compilers omit it) and simply forcing "/tmp".  (It
#ifdef's a wild range of targets -- compared to some of them, MS-DOS is
practically a Unix clone -- so I'm having a little difficulty being sure
of just what the code _is_ doing.)

  ***********

[Making a special case for MinGW in the #defines]

[Recompiling it]

[Yup, runs OK under Cygwin]

[And....  runs OK on the MS-DOS prompt, too]

  ************

OK, it looks like I had a false alarm.  Apologies -- unfortunately, the
code did a damn good job of simulating a C runtime problem....

-- 
John W. Kennedy
(Hoping for enough disk space soon so that I can start
 moving thoroughly to Linux and have done with everything
 with the Evil Empire's slime on it.)

--
Want to unsubscribe from this list?
Check out: http://cygwin.com/ml/#unsubscribe-simple


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