summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* Correct the patch.dgp2016-06-161-4/+1
|
* Merge ec51e0603e. Segfaults again.dgp2016-06-163-29/+33
|\
| * [8ff0cb9fe1] Make Tcl_NREvalObj() (and friends) behave as documented, bydgp2013-08-213-29/+33
| |\ | | | | | | | | | only scheduling evaluation and not doing any of it until the caller routine returns. This fixes some serious errors in [coroutine] too.
| | * Tidy the code and add a test.dgp_purge_NRRunObjProcdgp2013-08-212-39/+9
| | |
| | * Don't allocate memory until you know you're going to use it and arrange fordgp2013-08-211-1/+2
| | | | | | | | | it to be freed. Leak!
| | * Don't use automatic storage to hold the invocation words of oo::define.dgp2013-08-211-5/+8
| | | | | | | | | That practice doesn't agree with NRE execution.
| | * merge trunkdgp2013-08-2112-29/+41
| | |\ | | |/ | |/|
| | * Push out a trial patch for more eyes to see.dgp2013-08-201-0/+22
| | |
| | * Testing doing away with the NRRunObjProc routine, which looks like adgp2013-08-191-0/+8
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | useless extra bounce on the NRE trampoline. Normal testing has no problem with it, but debug-enabled testing triggers an assert failure. Either it would be good to have a normal test that fails in the conditions of the assert failure, or it would be good to discover the assert is asserting something not actually required, and then make the purge.
* | | Merge up to a197e6853e. Still works. Next checkin restarts the segfaults.dgp2016-06-16360-10776/+18353
|\ \ \ | |/ /
| * | [3612422]: Refer to correct part of tclvars(n) rather than page itself.dkf2013-08-2112-29/+41
| |/
| * Make sure the errors raised by execution traces become errors raised by thedgp2013-08-152-3/+10
| |\ | | | | | | | | | traced command, as documented. Deletion of the traced command was supressing that.
| | * Make sure the errors raised by execution traces become errors raised bydgp2013-08-152-1/+8
| | | | | | | | | | | | the traced command, as documented. Deletion of the traced command was supressing that.
| * | The fix for [3610404] leads to a simplification in the implementation ofdgp2013-08-152-18/+1
| | | | | | | | | forward methods.
| * | Add several tests to check consistency of stack traces.dgp2013-08-142-0/+94
| | |
| * | [3610404] Re-resolution of command after enter traces invalidate epoch.dgp2013-08-142-0/+65
| |\ \ | | | | | | | | Make sure context is such that correct resolution happens.
| | * \ merge trunkbug_3610404dgp2013-08-14205-6902/+12206
| | |\ \ | | |/ / | |/| |
| * | | [a16752c252] Correct failure to call cmd deletion callbacks.dgp2013-08-143-26/+15
| |\ \ \ | | | |/ | | |/|
| | * | [a16752c252] Correct failure to call cmd deletion callbacks.dgp2013-08-143-26/+15
| | | |
| * | | Arrange for both execution traces and [info frame] to get their pre-substdgp2013-08-144-101/+70
| |\ \ \ | | | | | | | | | | source strings from a common routine.
| | * | | Consolidate some helper routines.dgp_bye_ctx_eval_flagdgp2013-08-143-29/+19
| | | | |
| | * | | Only schedule leave traces if enter traces complete successfully.dgp2013-08-141-12/+10
| | | | | | | | | | | | | | | | | | | | This avoids a memleak, and opens a data slot, so we can pass objc, objv without the need to copy them into a list value.
| | * | | Use a new flag value TCL_EVAL_SOURCE_IN_FRAME passed in by callers todgp2013-08-123-50/+14
| | | | | | | | | | | | | | | | | | | | | | | | | determine whether the pre-subst source information in a CmdFrame is to be used. This takes the place of numLevels cross checking, so that field is removed. Routines are consolidated as well.
| | * | | merge trunkdgp2013-08-121-1/+1
| | |\ \ \ | | |/ / / | |/| | |
| * | | | Never guess non-existing timezone name "America/Brasilia" on Windows. ↵jan.nijtmans2013-08-111-1/+1
| |\ \ \ \ | | | |/ / | | |/| | | | | | | Reported by Arnulf Wiedemann
| | * | | Never guess non-existing timezone name "America/Brasilia" on Windows. ↵jan.nijtmans2013-08-111-1/+1
| | | | | | | | | | | | | | | | | | | | Reported by Arnulf Wiedemann
| * | | | Many simplifications to the TIP 280 machinery.dgp2013-08-078-170/+39
| |\ \ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | * Removed support for non-NULL invoker when TCL_EVAL_DIRECT requested. * Eliminated TCL_EVAL_CTX eval flag. * Removed Tcl_Preserve-ability of ContLineLoc pointers. * Removed clLoc field of CompileEnv struct.
| * \ \ \ \ The value TCL_LOCATION_EVAL_LIST in the type field of a CmdFrame appears todgp2013-08-067-97/+59
| |\ \ \ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | exist only for the sake of taking great pains to make sure that pure list values remain pure list values. The value of pure list values is no longer what it once was. For a long long time now, any canonical list values have been equally good.
| | | | | * | Arrange for both execution traces and [info frame] to get their pre-substdgp2013-08-104-31/+59
| | | | | | | | | | | | | | | | | | | | | | | | | | | | source strings from a common routine, with care taken to reduce copying by that routine.
| | | | | * | Revised GetCommandSource() can (and thus should) return a normal zerodgp2013-08-091-2/+1
| | | | | | | | | | | | | | | | | | | | | refcount value.
| | | | | * | Revise GetCommandSource() to return a normal Tcl_Obj value.dgp2013-08-071-18/+8
| | | | | | |
| | | | | * | Give (objc, objv) their own ride from enter to leave traces.dgp2013-08-071-4/+4
| | | | |/ / | | | |/| |
| | | * | | Remove Tcl_Preserve support for ContLineLoc values. It's not needed.dgp2013-08-073-56/+5
| | | | | | | | | | | | | | | | | | This allows the clLoc field of CompileEnv struct to go away too.
| | | * | | Test for TclContinuationsGet() usage, and simplifications.dgp2013-08-072-17/+7
| | | | | |
| | | * | | Replace potentially memleak creating safety check of a "cannot happen"dgp2013-08-071-8/+6
| | | | | | | | | | | | | | | | | | condition with an assertion.
| | | * | | Add comment stating new limitation on Tcl(NR)EvalObjEx() interface.dgp2013-08-071-0/+5
| | | | | |
| | | * | | All use of the evalFlag value TCL_EVAL_CTX is unused by the code anddgp2013-08-062-88/+11
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | unreachable by extensions. This checkin removes all the code supporting that flag value. The consequence is that all the calls to TclNREvalObjEx() and its callers that are currently choosing not to pass the TCL_EVAL_DIRECT flag in when they pass in a non-NULL invoker will no longer be free to change their mind. That might be reason not to adopt this change.
| | | * | | Add assertions that will guide and protect more discovery of dead code for ↵dgp2013-08-062-3/+7
| | |/ / / | | | | | | | | | | | | | | | elimination.
| | * | | Drop TCL_LOCATION_EVAL_LIST now that it is unused.dgp_bye_location_eval_listdgp2013-08-061-4/+0
| | | | |
| | * | | Eliminate the union that is no longer needed.dgp2013-08-066-34/+26
| | | | |
| | * | | Revert the changes that went too far and broke things.dgp2013-08-061-13/+26
| | | | |
| | * | | The value TCL_LOCATION_EVAL_LIST in the type field of a CmdFrame appears todgp2013-08-064-87/+48
| |/ / / | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | exist only for the sake of taking great pains to make sure that pure list values remain pure list values. The value of pure list values is no longer what it once was. For a long long time now, any canonical list values have been equally good. This branch is Work In Progress eliminating the complication of the additional type value. Currently some minor botches are breaking execution tracing tests.
| * | | Mark unixforkevent-1.1 nonPortable, until proven on more platforms. jan.nijtmans2013-08-054-3/+11
| |\ \ \ | | |/ / | | | | Make sure that when testing for pthread_atfork the thread library is taken into account.
| | * | Mark unixforkevent-1.1 nonPortable, until proven on more platforms.jan.nijtmans2013-08-052-3/+3
| | | |
| * | | [3611643fff]: Support TclOO in autoload mechanism.dkf2013-08-032-0/+19
| | | |
| * | | Deal with the elaborate rip-apart-a-metaclass case as well.dkf2013-08-023-3/+30
| | | |
| * | | [9d61624b3d]: Stop crashes when emptying the superclass slot.dkf2013-08-022-19/+46
| | | |
| * | | [1905562] [8d2c0da36d] Raise the recursion limits on regexps to allowdgp2013-08-012-1/+4
| | | | | | | | | | | | | | | | existing regexps "in the wild" to continue working with Tcl 8.6. Latest example comes from DejaGnu.
| * | | Tcl_InitNotifier: Bug [a0bc856dcd]. Start notifier thread again if we were ↵jan.nijtmans2013-08-013-1/+163
| |\ \ \ | | |/ / | | | | | | | | forked, to solve Rivet bug 55153
| | * | Tcl_InitNotifier: Bug [a0bc856dcd]. jan.nijtmans2013-08-013-4/+166
| | |\ \ | | | | | | | | | | Start notifier thread again if we were forked, to solve Rivet bug 55153
generated by cgit v0.12 at 2024-07-15 17:23:03 (GMT)