These changesets represents the changes on the squid/tidyup_deferred_reads development branch of Squid. If there is any questions please contact squid-dev@squid-cache.org.
| Date | Author | Size | Description |
|---|
| 2005/12/13 07:16:04 | adri | +9 -3 | * start explicitly killing off the callback when the server side goes |
| 2005/11/04 03:49:10 | adri | +65 -66 | * Create a "HAVE_SELECT" define to be used in comm_select.c |
| 2005/11/03 09:14:53 | adri | +415 -19 | Bring over Stevens' epoll code, sans the deferred stuff. |
| 2005/11/02 02:33:44 | adri | +2 -2 | We don't need to know about this debugging anymore. |
| 2005/11/02 02:27:51 | adri | +2 -11 | Remove commSetDefer() |
| 2005/11/02 02:25:33 | adri | +2 -121 | * Remove the deferred read checking code for both poll and select. |
| 2005/11/02 02:12:22 | adri | +1 -3 | Keep removing the deferred reads stuff |
| 2005/11/02 02:11:12 | adri | +1 -3 | Start stripping out the deferred handler check. |
| 2005/10/28 10:24:31 | adri | +15 -1 | The FTP data code does close the FD when the data connection is closed, |
| 2005/10/26 09:49:59 | adri | +4 -4 | Reverse some not-needed changes |
| 2005/10/26 05:08:07 | adri | +7 -22 | Reverse more unneeded changes! |
| 2005/10/26 04:49:31 | adri | +3 -99 | remove the comm_read() stuff, I didn't have to use it! |
| 2005/10/26 03:46:26 | adri | +11 -3 | Hah, this was easier than I thought. Just check in clientReadRequest() |
| 2005/10/26 03:00:13 | adri | +4 -1 | Pre-declarations, for Cygwin |
| 2005/10/26 02:55:34 | adri | +19 -8 | refactor the client side read IO scheduling stuff slightly in |
| 2005/10/25 09:23:37 | adri | +6 -6 | Oops! I had my logic reversed. |
| 2005/10/25 09:09:20 | adri | +14 -13 | Replace the accept defer handler with a 1 second rescheduling. |
| 2005/10/25 08:50:59 | adri | +1 -4 | remove unneeded static decl |
| 2005/10/25 08:40:57 | adri | +6 -34 | Back out my changes to the ssl code and remove the comm defer handler. |
| 2005/10/23 14:55:57 | adri | +17 -6 | Exam avoidance! |
| 2005/10/15 04:34:20 | adri | +4 -23 | Replace the transitory function calls with direct calls to our new |
| 2005/10/13 09:14:35 | adri | +76 -19 | An evil reworking of the half-closed connection support stuff. |
| 2005/10/12 08:19:14 | adri | +26 -7 | Refactor out the stuff playing with client side defers. |
| 2005/10/12 08:18:01 | adri | +2 -6 | Remove commSetDefer()s |
| 2005/10/12 07:53:33 | adri | +12 -53 | * make storeCheckDefer() global - its useful in some places |
| 2005/10/12 07:52:55 | adri | +1 -2 | Oops, remove call to fwdCheckDefer - we do'nt need it now |
| 2005/10/12 07:39:16 | adri | +49 -12 | Bring the WAIS code into the kicked IO framework |
| 2005/10/12 07:38:28 | adri | +2 -1 | .. and make sure the pending read is 0 when we start |
| 2005/10/12 07:37:09 | adri | +2 -1 | Just to bring it in line with http.c/ftp.c - preinitialise the IO |
| 2005/10/12 07:32:05 | adri | +2 -1 | Clear pending_read flag |
| 2005/10/12 07:21:37 | adri | +45 -7 | Modify gopher to use kicked io |
| 2005/10/12 07:20:55 | adri | +1 -2 | Take out an unused commSetDefer() call |
| 2005/10/12 05:31:01 | adri | +11 -8 | * add in some extra debugging |
| 2005/10/12 04:23:06 | adri | +41 -2 | more work towards bringing the FTP code into the new world order. |
| 2005/10/10 14:09:28 | adri | +20 -23 | Start breaking out the data IO path into our little hinting framework. |
| 2005/10/09 16:02:10 | adri | +7 -1 | A little paranoia: if the callback for a memobject changes, generally |
| 2005/10/09 15:47:14 | adri | +17 -3 | After wondering why the fetches were happening at full speed when |
| 2005/10/09 15:43:54 | adri | +3 -3 | change hex address debugging to be in line with the rest o'squid |
| 2005/10/09 10:18:35 | adri | +9 -2 | Modify the client to allow me to 'sleep(1)' before each read. |
| 2005/10/09 07:18:14 | adri | +17 -12 | * just so I can see whether I've 'messed up' my server-side code, |
| 2005/10/09 07:05:00 | adri | +5 -4 | * tweak the debugging a little |
| 2005/10/07 03:23:11 | adri | +42 -25 | * extend the STKICK callback now to take a kick type, allowing us to |
| 2005/10/06 03:14:58 | adri | +70 -11 | * disable some of the paranoia checking in my new code as I'm changing |
| 2005/10/05 14:54:00 | adri | +23 -7 | Break out the read scheduling and cancelling. This will become more |
| 2005/10/05 14:48:14 | adri | +141 -174 | Reverse my changes for now. They'll come back, but after my next move. |
| 2005/10/05 13:50:43 | adri | +131 -127 | * add a comment to point out what the code is doing |
| 2005/10/05 13:31:02 | adri | +1 -2 | more unused variables. |
| 2005/10/05 13:25:23 | adri | +1 -3 | Start removing unneeded variables. |
| 2005/10/05 12:14:48 | adri | +87 -82 | Attempt to refactor this code a little. All the code paths inside |
| 2005/10/05 07:44:30 | adri | +2 -1 | Add back missing return. The code now should do exactly what the |
| 2005/10/05 07:40:49 | adri | +16 -8 | Break out the code which skips whitespace in between http requests. |
| 2005/10/05 06:55:30 | adri | +33 -15 | Start refactoring the IO part of httpReadReply() so its not to |
| 2005/10/02 10:35:45 | adri | +24 -1 | Commit a little documentation - something i need to think about as |
| 2005/10/02 09:45:28 | adri | +79 -4 | (no idea why everything seems touched) |
| 2005/09/27 08:10:16 | adri | +99 -3 | Begin implementing a comm_read() to complement comm_write(). its not |
| 2005/09/27 06:32:20 | adri | +37 -38 | Stop wasting a call to mempoolalloc/free when the structure is only |