ID: 32861 User updated by: lew at mailduct dot com Reported By: lew at mailduct dot com -Status: Bogus +Status: Open Bug Type: Filesystem function related Operating System: FreeBSD 4.11-REL PHP Version: 4.3.10 New Comment:
After testing with the latest php5 release, this still has a problem. Let me try to be more succinct in summarizing... When operating with wrapper_type of 'plainfile' (such as when tailing one or more files), the tv_sec and tv_usec (timeout) are ignored... because stream_select insists on classifying FEOF as an important event. While this may be true for a socket event (socket close, for example... which is why we have socket_select instead), it is not true for a plainfile event. Ask yourself this question -- How would you make use of the 'timeout' parameter when reading from a plainfile wrapper? It would never come into play if FEOF awakens the stream_select call! The purpose of stream_select is to let you multiplex streams... so you can WAIT for an action to take place on a stream (or a timeout). If FEOF is allowed to awaken the select (rather than waiting for a timeout or for data to become available), then there's not much point to using that call... You're forcing it to do [expensive] polling instead of relying on the select() interrupt structure to do the work for you. My second argument is this: Let's assume you insist FEOF should awaken the select. If that's the case, then it still is not working correctly... because if I *add data* to a file after the select detected EOF, and then do a select again, it *still* thinks no data is available (and that we're still at EOF even though we're not). This is VERY broken. Previous Comments: ------------------------------------------------------------------------ [2005-05-27 00:04:54] lew at mailduct dot com I have now tested this using both PHP 4.3.11 and 5.0.4, and the bug remains. My same prior notes apply. Please review. ------------------------------------------------------------------------ [2005-04-28 07:45:59] lew at mailduct dot com I believe that under the OS and PHP version given in the headers of this report, the stream timeout function is not working when used with fgets as well. However, I will load the current PHP on the box and test again, to ensure I provide you with accurate and complete information and I'll open a new pr if necessary. Let's close this one as Bogus. ------------------------------------------------------------------------ [2005-04-28 00:35:16] [EMAIL PROTECTED] >From the stream_select() manual page: The streams listed in the read array will be watched to see if characters become available for reading (more precisely, to see if a read will not block - in particular, a stream resource is also ready on end-of-file, in which case an fread() will return a zero length string). ------------------------------------------------------------------------ [2005-04-28 00:05:08] [EMAIL PROTECTED] Does this happen with PHP 4.3.11 or 5.0.4 ? ------------------------------------------------------------------------ [2005-04-27 21:53:44] lew at mailduct dot com Description: ------------ Timeouts are ignored in stream_select. Instead, it always returns with "1" as the number of streams affected, ignoring any timeout seconds that are set. This is true regardless of stream_set_blocking settings. This may or may not be related to my reported FEOF problems: pr #25649 (Sep 2003) pr #32858 (Apr 2005) Reproduce code: --------------- $fp = fopen( '/var/log/maillog','r' ); ## Tried both ways, got same results #stream_set_blocking( $fp,TRUE ); #stream_set_blocking( $fp,FALSE ); $r = array( $fp ); while( TRUE ) { $n = stream_select( $r,$w=NULL,$e=NULL,30 ); if( $n===FALSE ) echo "stream_select = FALSE\n"; else echo "stream_select = $n\n"; echo fgets($fp); print( '.' ); } Expected result: ---------------- Assuming no other program is writing to /var/log/maillog, I expect to see a timeout condition every 30 seconds, as such: stream_select = FALSE Actual result: -------------- Instead, I see iterations of: stream_select = 1 as fast as the system can execute the loop. ------------------------------------------------------------------------ -- Edit this bug report at http://bugs.php.net/?id=32861&edit=1