Mail Archives: cygwin/1999/07/10/02:49:53
In a message dated 7/9/99 3:21:39 PM PST, khan AT xraylith DOT wisc DOT EDU writes:
> on a Win9x machine?
> John Fortin just reported a bizarre problem -- sometimes gcc will
> simply ignore certain include files
I must have kept my ranting about this within reasonable bounds. It may have
something to do with how many files are open. I've seen cases where copying
an include file into the first directory designated by -I solved the problem
of failing to read the contents of an include file, even though the file
should have been found via a later -I, and cases where a symlink didn't work
as well as a complete copy of the include file. Since W2K has become
available, I've found it (using NTFS file systems) a better way to run cygwin
than either W95 or NT4, although NT4 runs cygwin better on a FAT file system
than W2K does on a FAT32. It's helpful also to specify --host and --target
whenever it may be applicable, and not to mix compilers built for
i586-cygwin32 with i586-pc-cygwin32 snapshots. The most reliable way I've
found to make the transition is to build gcc once with --host=i586-cygwin32
--target=i586-pc-cygwin32, install the result, and use it to bootstrap a pure
i586-pc-cygwin32 compiler suite. One of the symptoms of not getting host and
target right is the explicit error message about not finding a standard
include file (or wanting to find it in /usr/include), but the nastier ones
are those where the include is satisfied yet the declaration in it are
ignored.
Along a similar line, I've got the old Unix Version 6 struct program
(translate very old-fashioned Fortran to ratfor) working, but on cygwin it
frequently gets its text buffer pointers out of whack, while it continues to
work with minor patchups even on pc-linux-gnu.
--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe AT sourceware DOT cygnus DOT com
- Raw text -