Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | COMPILE_DEBUG big: fix bug in stack verification for {*} | mig | 2013-01-17 | 1 | -2/+6 |
| | |||||
* | discouraging the compiler from re-reading *pc in the peephole loop, part2 ↵ | mig | 2013-01-12 | 1 | -6/+2 |
| | | | | (any diff?) | ||||
* | discouraging the compiler from re-reading *pc in the peephole loop | mig | 2013-01-12 | 1 | -7/+10 |
| | |||||
* | even better ... or so I hope: also inlining INST_PUSH1 in the peephole, ↵ | mig | 2013-01-12 | 1 | -53/+40 |
| | | | | checking for ISC after LOAD1 and PUSH1 | ||||
* | better comments | mig | 2013-01-11 | 1 | -2/+6 |
| | |||||
* | fix for consecutive ISC (produced by [while 1 {...}) | mig | 2013-01-11 | 1 | -21/+16 |
| | |||||
* | testing a cheaper(?) INST_START_COMMAND | mig | 2013-01-11 | 1 | -49/+48 |
| | |||||
* | Name functions according to 'what' instead of 'how' in the [tailcall] | mig | 2013-01-11 | 1 | -1/+1 |
| | | | machinery, in view of making public some parts of it. | ||||
* | tailcall now running in a simpler model, with no eval-flags and no nre-stack ↵ | mig | 2013-01-10 | 1 | -13/+7 |
| | | | | rewriting; yieldto also requires one fewer bounce. Mostly from mig-nre-mods | ||||
* | fix off-by-one error introduced in bd7d7a2061 | mig | 2013-01-10 | 1 | -1/+1 |
| | |||||
* | improved stack checking under TCL_COMPILE_DEBUG | mig | 2013-01-05 | 1 | -10/+9 |
| | |||||
* | fix the new assertion, it should not trigger when an expansion is in progress | mig | 2013-01-05 | 1 | -2/+5 |
| | |||||
* | Add an assertion that checks the stack depth under TCL_COMPILE_DEBUG ↵ | mig | 2013-01-05 | 1 | -2/+10 |
| | | | | *before* stack cleanup. It currently triggers at several spots in the testsuite! | ||||
* | Insure that PURIFY builds cannot exploit the Tcl stack to hide mem defects. | mig | 2013-01-04 | 1 | -1/+13 |
| | |||||
* | Added mechanism for pre-parsing built-in ensembles even when they are not going | dkf | 2013-01-03 | 1 | -0/+63 |
|\ | | | | | to generate specific code. This provides a good speedup in some cases. | ||||
| * | Passing more tests. | dkf | 2013-01-02 | 1 | -9/+15 |
| | | |||||
| * | Working towards more efficient treatment of non-bytecoded ensemble subcommands. | dkf | 2012-12-31 | 1 | -0/+57 |
| | | |||||
* | | remove stray calls to Tcl_Alloc and friends: the core should only use ↵ | mig | 2013-01-02 | 1 | -3/+3 |
|/ | | | | ckalloc to allow MEM_DEBUG to work properly | ||||
* | [Bug 3588366]: Corrected implementation of bounds restriction for end-indexed | dkf | 2012-11-19 | 1 | -3/+0 |
| | | | compiled [string range]. Thanks to Emiliano Gavilan for diagnosis and fix. | ||||
* | re-enable bcc-tailcall, after fixing an infinite loop in the ↵ | mig | 2012-11-10 | 1 | -1/+1 |
| | | | | TCL_COMPILE_DEBUG mode | ||||
* | Added compilation of [array exists], [array set] and [array unset]. Fixed a ↵merge_to_trunk | dkf | 2012-11-05 | 1 | -6/+108 |
| | | | | whole bunch of issues with opcode issuing that were causing problems with stack depth calculations. | ||||
* | Added compilation of [string last] and improved the compilation of [string ↵ | dkf | 2012-11-03 | 1 | -1/+47 |
| | | | | range]. This in turn enables compilation of [namespace qualifiers] and [namespace tail] (also done). | ||||
* | Added compilation of [info object isa object] (i.e., object verification). | dkf | 2012-11-03 | 1 | -0/+5 |
| | |||||
* | Added more TclOO introspection bytecodes ([info object class], [info object ↵ | dkf | 2012-11-02 | 1 | -0/+28 |
| | | | | namespace]). Also moved TclOO-in-8.6 to using the main Tcl internal ensemble builder. | ||||
* | Added compilation of [tailcall]. Not a particularly efficient compilation ↵ | dkf | 2012-11-01 | 1 | -0/+49 |
| | | | | though; it does not detect tailcall-of-self as a special case. | ||||
* | Added compilation of [dict create] and [dict merge]. | dkf | 2012-11-01 | 1 | -2/+13 |
| | |||||
* | Added [dict exists] compilation; implementation is 95% shared with [dict get]. | dkf | 2012-10-30 | 1 | -2/+23 |
| | |||||
* | Some corrections and performance tweaks | dkf | 2012-10-30 | 1 | -46/+56 |
| | |||||
* | Compilation of [string first] and [string range] (with constant indices). | dkf | 2012-10-30 | 1 | -1/+66 |
| | |||||
* | Added compilation of simplest practical case of [string map]. | dkf | 2012-10-29 | 1 | -0/+52 |
| | |||||
* | fix INST_YIELD so that it worksdkf_bytecode_8_6_yield | mig | 2012-10-28 | 1 | -5/+21 |
| | |||||
* | Working towards a BCCed [yield]; this doesn't work right now. | dkf | 2012-10-26 | 1 | -0/+26 |
| | |||||
* | Compile [namespace which -command]; big performance saving in some contexts. | dkf | 2012-10-26 | 1 | -0/+10 |
| | |||||
* | yet another small introspector: [self] | dkf | 2012-10-19 | 1 | -0/+24 |
| | |||||
* | merge trunk | dkf | 2012-08-04 | 1 | -18/+20 |
|\ | |||||
| * | converting to using Tcl_Obj API for error message generation; part done | dkf | 2012-08-03 | 1 | -18/+20 |
| | | |||||
* | | Compilation of misc info sometimes used in high-performance code. | dkf | 2012-03-04 | 1 | -0/+64 |
|/ | |||||
* | * generic/tclExecute.c (INST_LIST_RANGE_IMM): Enhance implementation | dkf | 2012-02-16 | 1 | -2/+23 |
| | | | | so that shortening a (not multiply-referenced) list by lopping the end off with [lrange] or [lreplace] is efficient. | ||||
* | Make some of the logic in TclCompileObj less heavily nested, to improve clarity. | dkf | 2011-11-22 | 1 | -99/+98 |
| | |||||
* | Fix a bug where global precompiled code (A) called from a precompiled | andreask | 2011-11-21 | 1 | -7/+9 |
| | | | | | | | | | | | | | | | | | | | | procedure causes the core to recompile (A), triggering the trap laid inside, i.e. 'error "called a copy of compiled code"'. <pre>Example ----------------------ex1.tcl proc init {} { source ex2.tcl } init ----------------------ex2.tcl puts a ----------------------</pre> When run as precompiled code the 'puts a' is not executed, only the trap. Fixed by enclosing the offending code into a guard which prevents its execution for precompiled code. The change passes the entire testsuite. | ||||
* | Added support for having the dict var itself referenced by LVT index. | dkf | 2011-10-03 | 1 | -13/+47 |
| | |||||
* | Experimental compilation of the [dict with] subcommand. No tests yet, and not | dkf | 2011-10-02 | 1 | -4/+41 |
| | | | | yet certain that the added bytecode opcodes are correct; evaluation is still needed (but the test suite does pass...) | ||||
* | [Bug 2919042] Restore "valgrindability" of Tcl that was lost by the ↵ | ferrieux | 2011-08-09 | 1 | -4/+8 |
| | | | | streamlining of [exit], by conditionally forcing a full Finalize: use -DPURIFY or ::env(TCL_FINALIZE_ON_EXIT) | ||||
* | platform portable type matching in debug prints | dgp | 2011-07-12 | 1 | -1/+1 |
| | |||||
* | Correct cast for CURR_DEPTH to silence compiler warning. [Bug 3339502] | Joe Mistachkin | 2011-07-11 | 1 | -1/+1 |
| | |||||
* | Reverted the fix for [Bug 3274728] committed on 2011-04-06 (rev ↵ | andreask | 2011-06-08 | 1 | -3/+3 |
| | | | | [caf317ab68]) and replaced with one which is 64bit-safe. The existing fix crashed tclsh on Windows 64bit. | ||||
* | TclFreeIntRep() cleanup. | dgp | 2011-04-27 | 1 | -1/+0 |
| | |||||
* | * generic/tclExecute.c (TclCompileObj): earlier return if Tip280 gymnastics ↵ | mig | 2011-04-07 | 1 | -1/+5 |
|\ | | | | | | | not needed. | ||||
| * | * generic/tclExecute.c (TclCompEvalObj): earlier return if Tip280 gymnastics ↵ | mig | 2011-04-07 | 1 | -39/+37 |
| | | | | | | | | not needed | ||||
* | | last bugfix was incomplete | mig | 2011-04-07 | 1 | -1/+1 |
| | |