Re: new module 'freadseek'

[Eric Blake]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

According to Bruno Haible on 2/29/2008 2:29 AM:
| Here's the side-effecting companion of freadptr.

|   /* Increment the in-memory pointer.  This is very cheap (no system
calls).  */
|   buffered = freadahead (fp);
|   if (buffered > 0)
|     {
|       size_t increment = (buffered < offset ? buffered : offset);
| #elif defined __sferror             /* FreeBSD, NetBSD, OpenBSD, MacOS
X, Cygwin */
|       fp->_p += increment;
|       fp->_r -= increment;

test-freadseek is failing on cygwin:

82        ungetc ('@', stdin);
(gdb) s
83        ASSERT (freadseek (stdin, nbytes6) == 0);
(gdb) s
freadseek (fp=0x61103080, offset=9) at ../../gllib/freadseek.c:33
33        if (offset == 0)
(gdb) p *fp
$1 = {_p = 0x611030c6 "@", _r = 1, _w = 0, _flags = -31612, _file = 0, _bf = {
~    _base = 0x660248 "#!/bin/sh\n\n./test-freadseek${EXEEXT} 5 19 6 7 18 9
19 < \"$srcdir/test-freadseek.sh\" || exit 1\ncat
\"$srcdir/test-freadseek.sh\" | ./test-freadseek${EXEEXT} 5 19 6 7 18 9 19
|| exit 1\nexit 0\n", _size = 65536},
~  _lbfsize = 0, _data = 0x0, _cookie = 0x61103080,
~  _read = 0x610f83f0 <__sread>, _write = 0x610f86c0 <__swrite64>,
~  _seek = 0x610f85b0 <__sseek>, _close = 0x610f8620 <__sclose>, _ub = {
~    _base = 0x611030c4 "", _size = 3},
~  _up = 0x660280 "\"$srcdir/test-freadseek.sh\" || exit 1\ncat
\"$srcdir/test-freadseek.sh\" | ./test-freadseek${EXEEXT} 5 19 6 7 18 9 19
|| exit 1\nexit 0\n", _ur = 132, _ubuf = "\000\000@", _nbuf = "", _lb =
{_base = 0x0, _size = 0},
~  _blksize = 65536, _flags2 = 0, _offset = 188,
~  _seek64 = 0x610f8650 <__sseek64>, _lock = 0x12}

...
81        fd = fileno (fp);
(gdb)
82        if (fd >= 0 && lseek (fd, 0, SEEK_CUR) >= 0)
(gdb)
85            return fseek (fp, offset, SEEK_CUR);
(gdb) p offset
$4 = 8
(gdb) fin
Run till exit from #0  freadseek (fp=0x61103080, offset=8)
Value returned is $5 = -1


Somehow, the tweaks you did to the raw _fp to account for the 1-byte
readahead buffer from the arbitrary ungetc trip up the subsequent fseek.
I'm in the process of compiling a debug version of cygwin to try to find
out more details why cygwin's fseek failed, but it may take a while.

- --
Don't work too hard, make some time for fun as well!

Eric Blake             address@hidden
-----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

iD8DBQFHyZYN84KuGfSFAYARAgDmAJ9VcZkZfahfwz/hCd9nUtxKLzEFcwCfWH51
0jKsFINsyyjibHDpz79t+z8=
=kw0g
-----END PGP SIGNATURE-----