-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 According to Bruno Haible on 4/13/2007 4:15 AM: > Eric Blake wrote: >> I'm still worried, though, that after using >> rpl_fflush/rpl_fseek, will fread/fgetc pick up at the correct character? >> I hope not to rewrite the entire stdio. > > This should be feasible, since a stream has a state when the buffer is empty > and the file descriptor is not block-aligned: namely, early after program > initialization. If you ensure that all internal pointers/counters are set > according to this state, I don't expect that you would need to override > fread(), fgetc(), etc.
OK, looking further at the newlib and MacOS sources (which I guess means looking at the entire BSD family), I may have a solution that doesn't even require rpl_fseek. It looks like the optimized seek is bypassed if fp->_seek is not __sseek, after possibly calling it once for SEEK_CUR. So I think what I do is make rpl_fflush check if fp->_seek is __sseek, in which case I replace it with rpl_sseek. rpl_sseek is written to call __sseek, and to restore __sseek to fp->_seek if whence is not SEEK_CUR. Then, fseek immediately following will be non-optimized, because it will see the non-__sseek, but the optimization is restored following the fseek. On cygwin, __sseek is not exported, so I can't do a configure time link check for its existence. Instead, rpl_sseek will have to be dropped in for any non-NULL fp->_seek, and rpl_sseek is a bit more complicated because it has to maintain a mapping of which _seek to use and restore based on which fp (aka cookie) is passed in. - -- Don't work too hard, make some time for fun as well! Eric Blake [EMAIL PROTECTED] -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (Cygwin) Comment: Public key at home.comcast.net/~ericblake/eblake.gpg Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFGH4Tf84KuGfSFAYARAoiPAJ95tus/XB5DW+Ho8cR0RttnQfdPKQCg0Tte ghsZLsZE6CpK+zU/vZ5hIC0= =OcqB -----END PGP SIGNATURE-----
