| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
| |
* win/tclWinPort.h:
* win/tclWinSerial.c: added TIP #35 Windows enhancements for
serial configuration. [Patch #438509] (schroedter)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
80x performance hit for file I/O on Win* systems. On my system
it was closer to a 120x hit. Problem report by Uwe Traum <no
email address available>.
The fix goes like this: The obstacle is 'FlushFileBuffers',
executed whenever Tcl writes data to the OS, as Tcl has to wait
for the disk to complete I/O, and disks are slow. We remove that
obstacle. This opens another problem, [file size] reports back
wrong numbers. So for [file size] we add the call back in. As
optimization we keep track of the channels which were written to
and flush only these.
* win/tclWinFile.c (TclpObjStat): Added a call to
'TclWinFlushDirtyChannels'. This ensures that [file size] and
related commands report the correct size of a file even if Tcl
has recently written to it. Unixoid OS's always report the
correct size even for files with pending data, but Win*
syssystem don't. They only report what is actually on disk.
* win/tclWinInt.h: Added declaration of
'TclWinFlushDirtyChannels', making it available to other parts
of the tcl core.
* win/tclWinChan.c (TclWinFlushDirtyChannels): New, internal,
procedure. Goes through the list of open file channels and
forces the OS to flush its file buffers for all which were
written to since the last call of this function. This is an
expensive operation as Tcl has to wait for the OS to complete
actual writes to the disk.
(FileInfo): Added dirty flag required by the procedure above.
(FileOutputProc): Removed flushing of file buffers, setting the
dirty flag instead. This means that the previously incurred
delays do not happen anymore.
(TclWinOpenFileChannel): Added initialization of 'dirty' flag.
|
| | |
|
| |
|
|
|
|
|
| |
* win/tclWinChan.c: moved Win2K bug case test with GetStdHandle()
from TclpGetDefaultStdChannel into Tcl_MakeFileChannel to enable
a more general method in detecting invalid OS handles rather than
just a specific known case. [BUG: 5971]
|
| |
|
|
| |
8.4a2 code base, merged in with some existing new 8.4a2 features.
|
| |
|
|
|
| |
0x00010001 instead of INVALID_HANDLE_VALUE for WinMain apps.
Added a new test case to catch it.
|
| |
|
|
|
|
|
|
|
|
|
| |
* library/reg1.0/pkgIndex.tcl:
* win/tclWinChan.c:
* win/tclWinThrd.c: converted CRLF to LF the */tcl.hpj.in files
were not converted, as it confuses hcw locally. [Bug: 5096]
* win/Makefile.in: expanded cleanup or help files
* doc/Thread.3: minor macro cleanup
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* win/tclWinChan.c:
* win/tclWinFCmd.c:
* win/tclWinInit.c:
* win/tclWinPipe.c:
* win/tclWinSock.c: removed all code that supported Win32s. It
was no longer officially supported, and likely didn't work anyway.
* win/makefile.vc: removed 16 bit stuff, cleaned up.
* win/tcl16.rc:
* win/tclWin16.c:
* win/winDumpExts.c: these files have been removed from the
source tree (no longer necessary to build)
* win/aclocal.m4: made it just include tcl.m4
|
| |
|
|
|
|
| |
channels. Checking for bad/unusable std channels was moved to Tk
since its only purpose was to check whether to use the Tk Console
Window for the std channels. [Bug: 2393 2392 2209 2458]
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
