summaryrefslogtreecommitdiffstats
path: root/generic
Commit message (Collapse)AuthorAgeFilesLines
* Add some comments about possible other self-restacking troubles.dgp2014-05-021-1/+8
|
* Fully restore topChan resetting to accommodate self-restacking channels.dgp2014-05-021-17/+26
|
* Re-enable buffer recycling.dgp2014-05-021-1/+1
|
* Stop the segfault in iogt-2.4. First by changing the UpdateInterest()dgp2014-05-011-1/+8
| | | | | | | call that triggers it. "downChanPtr" may no longer be the right argument at that point. Second, after ending the segfault, the test became an infinite loop (nested unstacking?! whoa.), so revised the test to one that terminates (and passes). Left behind a comment that the recursive unstacking case may require more examination.
* merge trunkdgp2014-05-013-13/+61
|\
| * We must Preserve channels if we're going to use TclChanCaughtErrorBypass() ↵dgp2014-05-011-3/+22
| |\ | | | | | | | | | to get error information after channel routines are called (and have possibly called for the channel to go away).
| | * We must Preserve channels if we're going to use TclChanCaughtErrorBypass()dgp2014-05-011-4/+22
| | | | | | | | | | | | to get error information after channel routines are called (and have possibly called for the channel to go away).
| | * Fix more corner-cases like ↵jan.nijtmans2014-05-014-12/+30
| | | | | | | | | | | | [0e92c404f19ede5b2eb06e6db27647d3138cc56|0e92c404f1]: The only place where a type of &tclByteArrayType can be trusted is when determining its length, because the character length of a (UTF-8) string is always equal to the byte length of the byte array.
| | |
| | \
| *-. \ Merge in the fixes for improper ChannelBuffer sharing via recycling.dgp2014-05-011-8/+27
| |\ \ \ | | | |/
| | | * Better (safer) fix for [0e92c404f1]jan.nijtmans2014-04-301-1/+1
| | |/
| | * Stop the segfaults in [close] during [gets] tests.dgp2014-04-301-9/+14
| | | | | | | | | Not sure this is the right behavior, but it's better than crashing.
| | * Panic message to pinpoint the cause of iocmd-21.23 segfault.dgp2014-04-301-0/+3
| | |
| | * Make sure no shared ChannelBuffers get recycled.dgp2014-04-291-0/+11
| | |
| * | [82e7f67325] Fix an evil refcount problem in compiled [string replace].dkf2014-04-301-2/+12
| | |
* | | merge trunkdgp2014-04-291-2/+4
|\ \ \ | |/ /
| * | Fix bug exposed when buffer recycling is disabled.dgp2014-04-291-2/+4
| |\ \ | | |/
| | * Fix bug exposed when buffer recycling is disabled.dgp2014-04-291-2/+4
| | |\
| | | * Revise the logic for setting TCL_ENCODING_END in the outputEncodingFlagsdgp_no_buffer_recycledgp2014-04-291-2/+4
| | | | | | | | | | | | so it does not rely on buffer recycling.
| | | * Disable buffer recycling to expose bugs for fixing.dgp2014-04-251-1/+1
| | |/
* | | Disable buffer recycling, which creates mysteries.dgp2014-04-251-1/+1
| | |
* | | Test iortrans-4.8.2 demos an infinite loop. Possible trouble withdgp2014-04-251-0/+5
|/ / | | | | pushback buffers.
* | Make sure the ReflectedChannel struct is freed in the handler thread,dgp2014-04-241-2/+2
| | | | | | | | where it was allocated. This constraint allows the struct to safely hold Tcl_Obj values, which has been convenient for storing callback commands.
* | [3493120] Plug memory leak in thread exit.dgp2014-04-243-2/+34
|\ \ | |/
| * [3493120] Plug memory leak in thread exit.dgp2014-04-233-2/+34
| |\
| | * Memory leak after thread exit, fixed (alloc cache released by exit), belong ↵sebres2014-04-223-2/+34
| | | | | | | | | | | | | | | | | | | | | | | | to ticket [3493120] Moved over to branch bug-3493120. This is not ready for the core-8-5-branch. Segfaults all over the place in a thread-enabled build on a CentOS system.
* | | Merge the [testchannel transform] fixes.dgp2014-04-221-47/+75
|\ \ \ | |/ /
| * | Add refcounting and preservation to [testchannel transform] to stop segfaultdgp2014-04-221-47/+75
| |/ | | | | in test iogt-2.4.
* | Merge refcounting machinery for ChannelBuffer.dgp2014-04-213-8/+46
|\ \ | |/
| * Added a refcounting mechanism to ChannelBuffers. Other edits todgp2014-04-213-8/+46
| | | | | | stop segfaults in tests iocmd-21.2[12].
* | Merge reflected channel improvements.dgp2014-04-171-95/+50
|\ \ | |/
| * Reflected channels. Keep a set of method names cached so we don't createdgp2014-04-171-40/+43
| | | | | | new each operation, and we can benefit from any lookup info cached in intreps.
| * Simplify reflected channels.dgp2014-04-171-55/+7
| | | | | | | | | | | | | | Instead of having two modes of Close operations and the need to choose between them with a special value of the methods field, when the initialize pass fails for some reason, simply do not create the channel so there's nothing that needs closing. Then the methods field no longer holds anything used, so eliminate it. All the methods checking is done by [chan create].
* | [d19a30db57] Fixed segfault in [clock] internals. Includes test.dgp2014-04-161-15/+21
|\ \ | |/
| * merge fork: Segmentation fault using some functions of tcl::clock, fixed, ↵jan.nijtmans2014-04-161-18/+21
| |\ | | | | | | | | | belong to ticket [d19a30db57]
| | * Segmentation fault using some functions of tcl::clock, fixed, belong to ↵sebres2014-04-161-18/+21
| | | | | | | | | | | | ticket [d19a30db57]
* | | Fix compiler warnings in win32/cygwin buildjan.nijtmans2014-04-162-6/+14
|\ \ \ | |/ /
| * | Fix compiler warnings in win32/cygwin build.jan.nijtmans2014-04-162-6/+14
| |/
* | Remove unused variable, don't use deprecated function, some formatting.jan.nijtmans2014-04-161-1/+1
| |
* | Upgrade from Winsock 1.1 to Winsock 2.2, which is always available on ↵jan.nijtmans2014-04-161-0/+6
|\ \ | |/ | | | | | | | | Win2000+. See: [http://msdn.microsoft.com/en-us/library/windows/desktop/ms742213%28v=vs.85%29.aspx] for details. Move winsock initialization to TclpInitPlatform(void), so we can be sure everywhere that we have an initialized winsock2. Stub entries for TclWinGetServByName/TclWinGetSockOpt/TclWinSetSockOpt are no longer necessary (will be removed in 9.0, but are kept in 8.x)
| * Upgrade from Winsock 1.1 to Winsock 2.2, which is always available on ↵jan.nijtmans2014-04-161-0/+6
| | | | | | | | | | | | Win2000+. See: [http://msdn.microsoft.com/en-us/library/windows/desktop/ms742213%28v=vs.85%29.aspx] for details. Move winsock initialization to TclpInitPlatform(void), so we can be sure everywhere that we have an initialized winsock2. Stub entries for TclWinGetServByName/TclWinGetSockOpt/TclWinSetSockOpt are no longer necessary (will be removed in 9.0, but are kept in 8.x)
* | Purge the (now unused) argc and argv fields.dgp2014-04-151-124/+4
|\ \ | |/
| * Purge (now unused) argc and argv fields.dgp2014-04-151-107/+3
| |
* | [88aef05cda] Stop reentrancy segfault in reflected channels by managingdgp2014-04-151-2/+20
|\ \ | |/ | | | | callbacks as (copies of) lists, not shared Tcl_Obj arrays. Still could use cleanup and improvements.
| * [88aef05cda] Stop reentrancy segfault in reflected channels by managingdgp2014-04-151-2/+20
| | | | | | | | callbacks as (copies of) lists, not shared Tcl_Obj arrays. Still could use cleanup and improvements.
* | [e663138a06] Fix the new INST_NUM_TYPE instruction so that the boundarydgp2014-04-141-6/+22
| | | | | | | | cases of [string is] on integral values are computed right. Code is now correct, though still suffers from a large amount of ugly.
* | Fix bug [e663138a06]: Test failures in "string is"jan.nijtmans2014-04-101-0/+9
| |
* | Fix for Bug [97069ea11a]: Set return message in close if a flush error is ↵oehhar2014-04-041-1/+11
|\ \ | |/ | | | | reported (which may be an error from a background flush)
| * Fix test failure socket-2.9: "1 {not owner}" instead of "1 {couldn't open ↵oehhar2014-04-011-9/+8
| | | | | | | | socket address already in use}" by only setting returned error message if not jet set.
| * Set return message in close if a flush error is reported (which may be an ↵oehhar2014-04-011-2/+13
| | | | | | | | error from a background flush). Ticket [97069ea11a]
* | Merge minimal fix for iocmd-23.11.dgp2014-03-282-10/+30
|\ \ | |/ | | | | | | Note that top channel regeneration is removed, so that Preserve/Release call pairs are sure to operate on the same pointers. Other bug fixes may need to change that.