| Commit message (Collapse) | Author | Age | Files | Lines |
|\ |
|
| |\ |
|
| | |
| | |
| | | |
rest of Tcl source code. No ABI change. API change *should* be harmless.
|
| | |
| | |
| | | |
the Engineering Manual.
|
|\ \ \
| |/ /
| | | |
cause more harm than good. Purged them (except in zlib files).
|
| |\ \
| | |/
| | | |
more harm than good. Purged them.
|
| | |
| | |
| | | |
more harm than good. Purged them.
|
| | |
| | |
| | |
| | |
| | |
| | | |
* tests/io.test: which used the number of _written_ bytes or
character to update the counters for the read bytes/characters.
New test io-53.11. This is a backward port from the 8.5 branch.
|
| | |
| | |
| | |
| | |
| | |
| | | |
0-length writes. When closing pipes which have already been closed
not skipping leads to spurious SIG_PIPE signals. Reported by
Mikhail Teterin <mi+thun@aldan.algebra.com>.
|
| | |
| | |
| | |
| | | |
something that doesn't crash the test suite.
|
| | |
| | |
| | |
| | |
| | |
| | | |
Alexandre Ferrieux's patch for [Bug 2270477] to prevent infinite
looping during finalization of channels not bound to
interpreters.
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Alexandre Ferrieux <ferrieux@users.sourceforge.net> to fix the
[Bug 1965787]. 'tell' now works for locations > 2 GB as well
instead of going negative.
* generic/tclIO.c (Tcl_SetChannelBufferSize): Accepted a patch by
* tests/io.test: Alexandre Ferrieux <ferrieux@users.sourceforge.net>
to fix the [Bug 1969953]. Buffersize outside of the supported
range are now clipped to nearest boundary instead of ignored.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* io.test (io-53.8a): Ferrieux <ferrieux@users.sourceforge.net>,
to shift EOF handling to the async part of the command if a
callback is specified, should the channel be at EOF already when
fcopy is called. Testcase by myself.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* generic/tclIO.c: Additional changes to data structures for fcopy
* generic/tclIO.h: and channels to perform proper cleanup in case
of a channel having two background copy operations running as is
now possible.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
TclCopyChannel): New macro, and the places using it. This change
allows for bi-directional fcopy on channels. Thanks to Alexandre
Ferrieux <ferrieux@users.sourceforge.net> for the patch.
* tests/io.test (io-53.9): Made test cleanup robust against the
possibility of slow process shutdown on Windows. Backported from
Kevin Kenny's change to the same test on the 8.5 and head
branches.
|
| | |
| | |
| | |
| | |
| | |
| | | |
* tests/io.test: prevent fcopy from calling -command synchronously
the first time. Thanks to Alexandre Ferrieux
<ferrieux@users.sourceforge.net> for report and patch.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
[Bug 780533], with many thanks to Alexandre Ferrieux
<ferrieux@users.sourceforge.net> for tracking it down and
providing a solution. Still have to convert his test script into
a proper test case.
|
| | | |
|
| | |
| | |
| | |
| | |
| | | |
aka coverity #51. Extended loop condition, added checking for
NULL to prevent seg.fault.
|
| | |
| | |
| | |
| | | |
inheritance to be #defined out (default remains in).
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
commentary to a piece of code which relies on BUFFER_PADDING to
create enough space at the beginning of each buffer forthe
insertion of partial multi-byte data at the beginning of a
buffer. To explain why this code is ok, and as precaution if
someone twiddled the BUFFER_PADDING into uselessness.
* generic/tclIO.c (ReadChars): [SF Tcl Bug 1462248]. Added code
temporarily suppress the use of TCL_ENCODING_END set when eof
was reached while the buffer we are converting is not truly the
last buffer in the queue. together with the Utf bug below it was
possible to completely bollox the buffer data structures,
eventually crashing Tcl.
* generic/tclEncoding.c (UtfToUtfProc): Fixed problem where the
function accessed memory beyond the end of the input
buffer. When TCL_ENCODING_END is set and the last bytes of the
buffer start a multi-byte sequence. This bug contributed to [SF
Tcl Bug 1462248].
|
| | | |
|
| | |
| | |
| | |
| | | |
* generic/tclIOUtil.c: interp == NULL arguments. [Bug 1380662]
|
| | |
| | |
| | |
| | | |
until we netrieve next statePtr from it.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
patch for [SF Tcl Bug 1359094]. This moves the retrieval of the
next channel state to the end of the loop, as the called
closeproc may close other channels, i.e. modify the list we are
iterating, invalidating any pointer retrieved earlier.
|
| | |
| | |
| | |
| | | |
for more info.
|
| | | |
|
| | |
| | |
| | |
| | | |
event processing when channel is being closed/cutted.
|
| | |
| | |
| | |
| | |
| | | |
timer for the channel. Also, prevents events still in
the event queue from triggering on the current channel.
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* tests/io.test: changed from ten bytes to one byte. Need
* tests/iogt.test: for this change was proven by
Ross Cartlidge <rossc@cisco.com> where [read stdin 1] was grabbing
10 bytes followed by starting a child process that was intended to
continue reading from stdin. Even with -buffersize set to one,
nine chars were getting lost by the buffersize over reading for
the native read() caused by [read].
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* generic/tclDecls.h: Regenerated from tcl.decls.
* generic/tclStubInit.c:
* doc/CrtChannel.3: Documentation of extended API,
* generic/tcl.decls: extended testsuite, and
* generic/tcl.h: implementation. Removal of old
* generic/tclIO.c: driver-specific TclpCut/Splice
* generic/tclInt.h: functions. Replaced with generic
* tests/io.test: thread-action calls through the
* unix/tclUnixChan.c: new hooks. Update of all builtin
* unix/tclUnixPipe.c: channel drivers to version 4.
* unix/tclUnixSock.c: Windows drivers extended to
* win/tclWinChan.c: manage thread state in a thread
* win/tclWinConsole.c: action handler.
* win/tclWinPipe.c:
* win/tclWinSerial.c:
* win/tclWinSock.c:
* mac/tclMacChan.c:
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* generic/tclIO.c (Tcl_UnregisterChannel): 'Tcl_Close' while the
* generic/tclIO.c (Tcl_Close): close callbacks are
run. Checked in 'Tcl_Close' and 'Tcl_Unregister' to prevent
recursive call of 'close' in the close-callbacks. This is a
possible error made by implementors of virtual filesystems based
on 'tclvfs', thinking that they have to close the channel in the
close handler for the filesystem.
* generic/tclIO.c:
* generic/tclIO.h:
* Not reverting, but #ifdef'ing the changes from May 19, 2004 out
of the core. This removes the ***POTENTIAL INCOMPATIBILITY***
for channel drivers it introduced. This has become possible due
to Expect gaining a BlockModeProc and now handling blockingg and
non-blocking modes correctly. Thus [SF Tcl Bug 943274] is still
fixed if a recent enough version of Expect is used.
* doc/CrtChannel.3: Added warning about usage of a channel without
a BlockModeProc.
|
| | |
| | |
| | |
| | | |
after shifting the code.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* tclIO.h: [SF Tcl Bug 462317], see ChangeLog entry
2001-09-26. The fix done at that time is incomplete. It
is possible to get around it if the actual read
operation is defered and not executed in the event
handler itself. Instead of tracking if we are in an
read caused by a synthesized fileevent we now track if
the OS has delivered a true event = actual data and
bypass the driver if a read finds that there is no
actual data waiting. The flag is cleared by a short or
full read.
|
| | |
| | |
| | |
| | |
| | | |
930851]. When changing the eofchar we have to zap the related
flags to prevent them from prematurely aborting the next read.
|
| | |
| | |
| | |
| | | |
[Bug 405995]
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
TCL_READABLE now dropping interest in TCL_EXCEPTION too. This
fixes a bug where Expect detects eof on a file prematurely on
solaris 2.6 and higher. A much more complete explanation is in
the code itself (40 lines of comments for a one-line change :)
|
| | |
| | |
| | |
| | |
| | |
| | | |
are problems than to try and compile everything manually on all the
platforms. This way the AS auto-build gets the change without fuss,
and I can read the logs tomorrow.
|
| | |
| | |
| | |
| | | |
callback asynchronously, even for size zero.
|
| | |
| | |
| | |
| | | |
by [fcopy] to exceed 2GB. Can happen when no -size parameter given.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
the EventScriptRecord (esPtr) fully before handing it to
Tcl_CreateChannelHandler for registration. Otherwise a reflected
channel calling 'chan postevent' (== Tcl_NotifyChannel) in its
'watchProc' will cause the function 'TclChannelEventScriptInvoker'
to be run on an uninitialized structure.
|
| | |
| | |
| | |
| | |
| | |
| | | |
which used the number of _written_ bytes or character to update
the counters for the read bytes/characters. See last entry for the
test case.
|
| | |
| | |
| | |
| | |
| | | |
saving the errno from the first of two FlushChannel()s. Uneasy to
test; might need specific channel drivers. Four-hands with aku.
|
| | |
| | |
| | |
| | |
| | |
| | | |
0-length writes. When closing pipes which have already been closed
not skipping leads to spurious SIG_PIPE signals. Reported by
Mikhail Teterin <mi+thun@aldan.algebra.com>.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
to permit reads to continue up to the string limits of Tcl values.
Before revisions, large read attempts could panic when as little as
half the limiting value length was reached. [Patch 2107634]
Thanks to Sean Morrison and Bob Parker for their roles in the fix.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* generic/tclPipe.c (FileForRedirect): for getHandleProc == NULL, this
is allowed. Provide a nice error message in the bypass area. Updated
caller to check the bypass for a mesage. Bug reported by Andy
Sonnenburg <andy22286@users.sourceforge.net>. Backported from CVS
head.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* tests/io.test: internal representation of the tclChannelType to
contain not only the ChannelState pointer, but also a reference to
the interpreter it was made in. Invalidate and recompute the
internal representation when it is used in a different interpreter
(Like cmdName intrep's). Added testcase. [Bug 2407783].
|