| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
| |
TclInvalidateStringRep() call when we directly manipulate the
intrep of an unshared "list" Tcl_Obj. [Bug 1672585].
|
| |
|
|
|
|
| |
to append Tcl_Obj's instead of strings. [RFE 1669420]
|
|
|
|
| |
optimizations.
|
| |
|
|
|
|
|
|
| |
to work, immediate operands referring to aux-data must be identified as such in the instruction descriptor table using OPERAND_AUX4 (all are always 4 bytes).
Rewrote the compiled [dict update] so that it stores critical non-varying data in an aux-data value instead of a (shimmerable) literal. [Bug 1671001]
|
|
|
|
| |
and optimizations.
|
|
|
|
| |
and optimizations.
|
| |
|
|
|
|
| |
spirit; avoid shimmer effects rather than react to them.
|
|
|
|
|
| |
* tests/foreach.test (foreach-1.14): useful error information when
loop variable sets fail.
|
|
|
|
|
|
| |
efficient private copy of the list argument, so we can operate on the
list elements directly with no fear of shimmering effects. Replaces
defensive coding schemes that are otherwise required.
|
|
|
|
|
| |
efficient private copies of the variable and value lists, so we can
operate on them without any special shimmer defense coding schemes.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
refcount release between NewRC and FreeRC for the channel handle
object, spotted by Don Porter. This fixes the bug 1667990.
|
|
|
|
|
| |
of the objv array that used to be a workaround for Bug 404865. That bug
is long fixed.
|
|
|
|
|
|
|
|
|
|
|
|
| |
the recounting logic of the List internal rep need not be repeated
there. Better encapsulation of internal details.
* generic/tclInt.h: New internal routine TclListObjCopy() used
* generic/tclListObj.c: to efficiently do the equivalent of
[lrange $list 0 end]. After some experience with this, might be
a good candidate for exposure as a public interface. It's useful
for callers of Tcl_ListObjGetElements() who want to control the
ongoing validity of the returned objv pointer.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
satisfying a require for the regular package, demonstrating a
corner case specified in TIP#280. More notes in the comments to
the test.
|
| |
|
| |
|
| |
|
|
|
|
| |
* unix/configure: autoconf-2.59
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
* tools/fix_tommath_h.tcl: Further tweaking for the x86-64.
The change is to make 'mp_digit' be an 'unsigned int' on that
platform; since we're using only 32 bits of it, there's no
reason to make it a 64-bit 'unsigned long.'
* generic/tclTomMath.h: Regenerated.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
__x86_64__ that caused Tommath to use __attributes(TI)__ for
the mp_word type. Tetra-int's simply fail on too many gcc-glibc-OS
combinations to be ready for shipment today, even if they
work for some of us. This change allows reversion of das's change
of 2006-08-18 that accomplised the same thing on Darwin.
[Bugs
* generic/tclTomMath.h: Regenerated.
|
|
|
|
|
|
|
|
|
| |
__x86_64__ that caused Tommath to use __attributes(TI)__ for
the mp_word type. Tetra-int's simply fail on too many gcc-glibc-OS
combinations to be ready for shipment today, even if they
work for some of us. This change allows reversion of das's change
of 2006-08-18 that accomplised the same thing on Darwin.
* generic/tclTomMath.h: Regenerated.
|
| |
|
|
|
|
|
| |
with unsigned size_t to correctly validate stackSize in the 2^31+
range.
|
|
|
|
|
|
| |
* tests/namespace.test: introduced in Patch 1577278 that caused
[namespace delete ::] to be effective only at level #0. New test
namespace-7.7 should prevent similar error in the future. [Bug 1655305]
|
| |
|
| |
|
| |
|
|
|
|
|
| |
* tests/namespace.test: TclMatchIsTrivial optimization on
[namespace children $namespace $pattern].
|
|
|
|
|
|
|
| |
available to define MODULE_SCOPE effective on all platforms.
* unix/configure.in: add caching to -pipe and zoneinfo checks.
* unix/configure: autoconf-2.59
* unix/tclConfig.h.in: autoheader-2.59
|
| |
|
|
|
|
|
| |
to cover common case where there's big savings over the fully general
path. Thanks to Peter MacDonald. [Bug 1626518].
|
|
|
|
| |
to Andy Goth [Bug 1602538].
|