On Wed, 29 Aug 2001 01:14:23 +1000 (EST), Brian Havard wrote:
>On 28 Aug 2001 01:56:09 -0000, [EMAIL PROTECTED] wrote:
>
>>wrowe 01/08/27 18:56:09
>>
>> Modified: file_io/win32 readwrite.c seek.c
>> Log:
>> Found a very ugly reaction to using apr_file_seek(APR_CUR, -value) in
>> conjuction with buffered reads. Thank you for toggling that case Jeff,
>> so I could shoot out this bug ;)
>
>Did this actually fix things for you? You see the new offset calculation in
>the buffered seek is wrong (my fault). I've only tested on OS/2 but it should
>apply equally to all platforms. The fix is:
Forget it, different bug. Just applied similar fix to the OS/2 version.
For unix, the patch below should do it but I can't test.
Index: readwrite.c
===================================================================
RCS file: /home/cvs/apr/file_io/unix/readwrite.c,v
retrieving revision 1.71
diff -u -r1.71 readwrite.c
--- readwrite.c 2001/08/24 12:19:01 1.71
+++ readwrite.c 2001/08/28 16:19:02
@@ -137,16 +137,17 @@
}
while (rv == 0 && size > 0) {
if (thefile->bufpos >= thefile->dataRead) {
- thefile->dataRead = read(thefile->filedes, thefile->buffer,
APR_FILE_BUFSIZE);
- if (thefile->dataRead == 0) {
+ int bytesread = read(thefile->filedes, thefile->buffer,
APR_FILE_BUFSIZE);
+ if (bytesread == 0) {
thefile->eof_hit = TRUE;
rv = APR_EOF;
break;
}
- else if (thefile->dataRead == -1) {
+ else if (bytesread == -1) {
rv = errno;
break;
}
+ thefile->dataRead = bytesread;
thefile->filePtr += thefile->dataRead;
thefile->bufpos = 0;
}
--
______________________________________________________________________________
| Brian Havard | "He is not the messiah! |
| [EMAIL PROTECTED] | He's a very naughty boy!" - Life of Brian |
------------------------------------------------------------------------------
