| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
| |
* doc/tclvars.n: Performance fix for TIP #280.
* generic/tclBasic.c:
* generic/tclExecute.c:
* generic/tclInt.h:
* generic/tclInterp.c:
* tests/info.test:
* tests/interp.test:
|
|
|
|
| |
revert tclInt.h to what it was before, and relax the relation between Tcl_CallFrame and CallFrame.
|
|
|
|
| |
and added test cases for it.
|
|
|
|
| |
evaluation.
|
|
|
|
|
|
| |
TCL_EVAL_DIRECT evaluation.
* tests/info.test: Resolve ambiguous resolution of variable "res".
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
EvalTokensStandard, Tcl_EvalEx, EvalEx, TclAdvanceContinuations,
TclEvalObjEx):
* generic/tclCmdMZ.c (Tcl_SwitchObjCmd, TclListLines):
* generic/tclCompCmds.c (*):
* generic/tclCompile.c (TclSetByteCodeFromAny, TclInitCompileEnv,
TclFreeCompileEnv, TclCompileScript):
* generic/tclCompile.h (CompileEnv):
* generic/tclInt.h (ContLineLoc, Interp):
* generic/tclObj.c (ThreadSpecificData, ContLineLocFree,
TclThreadFinalizeObjects, TclInitObjSubsystem,
TclContinuationsEnter, TclContinuationsEnterDerived,
TclContinuationsCopy, TclContinuationsGet, TclFreeObj):
* generic/tclParse.c (TclSubstTokens, Tcl_SubstObj):
* generic/tclProc.c (TclCreateProc):
* generic/tclVar.c (TclPtrSetVar):
* tests/info.test (info-30.0-24):
Extended parser, compiler, and execution with code and attendant
data structures tracking the positions of continuation lines which
are not visible in script Tcl_Obj*'s, to properly account for them
while counting lines for #280.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
TclArgumentBCRelease, TclArgumentGet):
* generic/tclCompile.c (EnterCmdWordIndex, TclCleanupByteCode,
TclInitCompileEnv, TclCompileScript):
* generic/tclCompile.h (ExtCmdLoc):
* generic/tclExecute.c (TclExecuteByteCode):
* generic/tclInt.h (ExtIndex, CFWordBC):
* tests/info.test (info-39.0):
Backport of some changes made to the Tcl head, to handle literal
sharing better. The code here is much simpler (trimmed down)
compared to the head as the 8.5 branch is not bytecode compiling
whole files, and doesn't compile eval'd code either.
Reworked the handling of literal command arguments in bytecode to
be saved (compiler) and used (execution) per command (See the
TCL_INVOKE_STK* instructions), and not per the whole bytecode.
This removes the problems with location data caused by literal
sharing in proc bodies. Simplified the associated datastructures
(ExtIndex is gone, as is the function EnterCmdWordIndex).
|
|
|
|
|
| |
* generic/tclBasic.c: 'proc', 'cmd' and 'inst' probes (does
_not_ require a platform with DTrace).
|
|
|
|
| |
TCL_ALLOW_EXCEPTIONS is tested.
|
|
|
|
|
|
|
| |
empty string as path (TclEvalEx). In 8.4 the missing code caused
panics in the testsuite. It doesn't in 8.5. I am guessing that the
code path with the missing the incr-refcount is not invoked any
longer. Because the bug in itself is certainly the same.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
of location for uplevel scripts.
* generic/tclCompile.c (TclInitCompileEnv): Reorganized the
initialization of the #280 location information to match the flow
in TclEvalObjEx to get more absolute contexts.
* generic/tclBasic.c (TclEvalObjEx): Moved the pure-list
optimization out of the eval-direct code path to be done always,
i.e. even when a compile is requested. This way we do not loose
the association between #280 location information and the list
elements, if any.
|
|
|
|
|
|
|
|
|
|
| |
* generic/tclCmdIL.c: immediately, without search. Reworked setup
* generic/tclCompile.c: of eoFramePtr, doesn't need the line
* tests/info.test: information, more sensible to have everything
on line 1 when eval'ing a pure list. Updated the users of the line
information to special case this based on the frame type (i.e.
TCL_LOCATION_EVAL_LIST). Added a testcase demonstrating the new
behaviour.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
* generic/tclCompile.c: for #280 to fix the abysmal performance
* generic/tclCompile.h: for deep recursion, replaced the linear
* generic/tclExecute.c: search through the whole stack with
* generic/tclInt.h: another hashtable and simplified the data
structure used by the compiler (array instead of hashtable).
Incidentially this also fixes the memory leak reported via [Bug
2024937].
|
|
|
|
|
|
|
|
|
|
|
| |
* generic/tclCmdAH.c: frame), added the ability to track the
* generic/tclCompCmds.c: absolute location of literal procedure
* generic/tclCompile.c: arguments, and making this information
* generic/tclCompile.h: available to uplevel, eval, and
* generic/tclInterp.c: siblings. This allows proper tracking of
* generic/tclInt.h: absolute location through custom (Tcl-coded)
* generic/tclNamesp.c: control structures based on uplevel, etc.
* generic/tclProc.c:
|
|
|
|
| |
from bad TclStackAlloc() alignment. Thanks George Peter Staplin.
|
|
|
|
|
|
| |
* tests/expr.test: function and the [::tcl::mathfunc::abs]
command do not return the value of -0, or equivalent values with
more alarming string reps like -1e-350. [Bug 1893815].
|
|
|
|
|
|
|
|
|
|
| |
refcount management of Tcl_AppendObjToErrorInfo to become that of
a conventional Consumer routine. This preserves the ease of use
for the overwhelming common callers who pass in a 0-count value,
but makes the proper call with a non-0-count value less surprising.
* generic/tclEvent.c (TclDefaultBgErrorHandlerObjCmd): Revised the
one caller within Tcl itself which passes a non-0-count value to
Tcl_AppendObjToErrorInfo().
|
|
|
|
|
|
| |
an enter trace deletes or changes the command, prompting a reparsing.
Don't let the second pass lose commandPtr value allocated during the
first pass.
|
| |
|
|
|
|
|
| |
* generic/tclCompExpr.c: of struct TclOpCmdClientData to
* generic/tclCompile.h: accomodate C++ compilers [Bug 1855644]
|
| |
|
| |
|
|
|
|
| |
TclMakeEnsemble to do most of the work.
|
|
|
|
| |
add comments
|
|
|
|
|
|
|
|
|
| |
* generic/tclInt.h:
* unix/tclUnixInit.c:
* unix/tclUnixThrd.c: Fix stack checking via workaround for bug in
glibc's pthread_attr_get_np, patch from [Bug 1815573]. Many thanks
to Sergei Golovan (aka Teo) for detecting the bug and helping
diagnose and develop the fix.
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
THIS PATCH WAS REVERTED: initial (mis)measurements overstated the
perfomance wins, which turn out to be tiny. Not worth the
complication.
|
|
|
|
|
|
|
|
|
| |
* generic/tclExecute.c: bit INTERP_RESULT_UNCLEAN: shortcut for
* generic/tclInt.h: Tcl_ResetResult for the "normal" case:
* generic/tclProc.c: TCL_OK, no return options, no errorCode
* generic/tclResult.c: nor errorInfo, return at normal level.
* generic/tclStubLib.c: [Patch 1830184]
* generic/tclUtil.c:
|
|
|
|
| |
objTypes.
|
| |
|
| |
|
|
|
|
|
|
| |
* unix/configure.in:
* unix/tclUnixInit.c: detect stack grwoth direction at compile
time, only fall to runtime detection when crosscompiling.
|
| |
|
|
|
|
|
|
|
|
|
| |
* generic/tclInt.h:
* unix/tclUnixInit.c:
* win/tclWin32Dll.c: restore simpler behaviour for stack checking,
not adaptive to stack size changes after a thread is
launched. Consensus is that "nobody does that", and so it is not
worth the cost.
|
|
|
|
|
|
|
|
|
|
|
| |
* generic/tclInt.h:
* unix/tclUnixInit.c:
* unix/tclUnixPort.h:
* win/tclWin32Dll.c: modify the stack checking algorithm to
recheck in case of failure. The working assumptions are now that
(a) a thread's stack is never moved, and (b) a thread's stack can
grow but not shrink. Port to windows - could be more efficient,
but is already cheaper than it was.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
* generic/tclBasic.c:
* generic/tclExecute.c:
* generic/tclInt.h:
* generic/tclUnixInit.c:
* generic/tclUnixPort.h: new fields in interp (ekeko!) to cache
TSD data that is accessed at each command invocation, access
macros to replace Tcl_AsyncReady and TclpCheckStackSpace by much
faster variants [Patch 1829248]
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* generic/tclCompile.h: tracing of proc and command entry &
* generic/tclBasic.c: return, bytecode execution, object
* generic/tclExecute.c: allocation and more; with essentially
* generic/tclInt.h: zero cost when tracing is inactive;
* generic/tclObj.c: enable with --enable-dtrace configure
* generic/tclProc.c: arg (disabled by default, will only
* unix/Makefile.in: enable if DTrace is present).
* unix/configure.in: [Patch 1793984]
* macosx/GNUmakefile: enable DTrace support.
* macosx/Tcl-Common.xcconfig:
* macosx/Tcl.xcodeproj/project.pbxproj:
* unix/configure: autoconf-2.59
* unix/tclConfig.h.in: autoheader-2.59
|
|
|
|
|
| |
* generic/tclExecute.c: -DTCL_GENERIC_ONLY configuration. [Bug 1264623]
* unix/Makefile.in:
|
|
|
|
|
| |
* generic/tclCompExpr.c: the fields of the OpCmdInfo and
* generic/tclCompile.h: TclOpCmdClientData structs.
|
|
|
|
|
|
| |
* tests/thread.test: ::errorInfo and ::errorCode so that string
* tests/trace.test: reported to variable traces are fully
qualified in agreement with Tcl 8.4 operations.
|