summaryrefslogtreecommitdiffstats
path: root/unix
Commit message (Collapse)AuthorAgeFilesLines
* (cherry-pick): HAVE_TM_GMTOFF detection doesn't work if CFLAGS contains ↵jan.nijtmans2020-10-062-2/+2
| | | | -Werror. Here's the fix.
* Many more internal master/slave -> parent/child renamingsjan.nijtmans2020-09-011-3/+3
|
* Fix for http bug c2dc1da315. Add tests. Add detail about -handler to ↵kjnash2020-08-241-2/+2
| | | | http(n). Bump version to 2.9.5.
* Bugfixes to the earlier fix of bug cb0373bb33, which broke HTTP/1.0 ↵kjnash2020-08-121-2/+2
| | | | transactions in which the server indicates neither a Content-Length nor that it will close the socket on completion ("Connection: close"). The HTTP/1.1 rule is that the response header "Connection", if absent, must default to "keep-alive"; but this rule does not apply to HTTP/1.0. Add test http11-3.4 and bump version to 2.9.4.
* Revert unintended ckalloc -> Tcl_Alloc change from previous commitjan.nijtmans2020-08-111-4/+4
|
* Code cleanup: less of use "register" keyword, and better use of typecastsjan.nijtmans2020-08-114-64/+76
|
* Eliminate many unnecessary type-casts, mostly (size_t) when value is already ↵jan.nijtmans2020-08-113-4/+4
| | | | size_t or int
* Fix for bug cb0373bb33. In HTTP/1.1 the response header "Connection", if ↵kjnash2020-07-241-2/+2
| | | | absent, must default to "keep-alive". Add test http11-1.13 and bump version to 2.9.3.
* Use "Global" in stead of "Master" internally, where "Master" is not used in ↵jan.nijtmans2020-07-201-26/+26
| | | | the Master-Slave paradigm and where this makes sense.
* Re-generate unix/tclConfig.h.injan.nijtmans2020-06-291-6/+20
|
* MacOSX: add some type-casts, improving C++ compatibility. Make sure that ↵jan.nijtmans2020-06-282-4/+0
| | | | strstr and strtoul() are not configured as invalid: Mac is more strict in exact function signatures. Backported from 8.7
* Add "make dist" target to Travis build, as regression test. This should ↵jan.nijtmans2020-06-261-1/+1
| | | | continue to work.
* Merge 8.5jan.nijtmans2020-06-254-2/+18
|\
| * One include too muchjan.nijtmans2020-06-242-2/+0
| |
| * re-generate unix/configurejan.nijtmans2020-06-241-0/+9
| |
| * Fix [9080d6a871]: Build fix for tcl on macOS on Apple Siliconjan.nijtmans2020-06-244-32/+19
| |
| * Add "manifest.uuid" to the distribution, so we can see exactly where "make ↵jan.nijtmans2020-06-242-10/+37
| | | | | | | | dist" came from. In case of "git", we can see that too
* | Bump http version to 2.9.2kjnash2020-06-191-2/+2
| |
* | Merge-mark 8.5 (TZ update already done here)jan.nijtmans2020-06-082-3/+3
|\ \ | |/ | | | | Don't try to compile with MINGW32_ in /unix directory. Minor comment fix.
* | Remove PROTO_FLAGS support from the Makefile, since it isn't used anywhere ↵jan.nijtmans2020-05-051-7/+2
| | | | | | | | any more.
* | Merge-mark 8.5. Remove some ARGSUSED usage.jan.nijtmans2020-04-173-9/+0
|\ \ | |/
| * Unbreak shared windows build. Remove some ARGSUSED usage.jan.nijtmans2020-04-172-10/+0
| |
| * Clean-up some unnecessary spacing.jan.nijtmans2020-04-176-9/+9
| |
* | Upgrade Xcode from 11.3 to 11.4. Let Travis test whether "make install" ↵jan.nijtmans2020-04-162-24/+24
|\ \ | | | | | | | | | works when the destination directory contains a space.
| * \ Merge 8.6jan.nijtmans2020-04-151-1/+1
| |\ \
| * | | More double-quotesjan.nijtmans2020-04-152-22/+22
| | | |
| * | | Try adding some quotesjan.nijtmans2020-04-142-2/+2
| | | |
* | | | Merge 8.5jan.nijtmans2020-04-151-1/+1
|\ \ \ \ | |_|/ / |/| | / | | |/ | |/|
* | | Merge 8.5. Failing tests highlight ticket [1b1f5f0b53].dgp2020-04-131-1/+1
|\ \ \ | |/ / | | / | |/ |/|
| * Guarantee TclNeedSpace and TclFindElement have common definition of whitespacedgp2020-04-091-1/+1
| | | | | | | | by having both call the same routine. Create a macro form to contain performance costs and adapt callers.
* | Fix determination of TCL_PACKAGE_PATH if it contains multiple directories, ↵jan.nijtmans2020-04-102-4/+4
| | | | | | | | and at least one of them contains a space.
* | re-generate UNIX configure scriptjan.nijtmans2020-04-101-4941/+13397
| |
* | Introduce MODULE_INSTALL_DIR in make/nmake files, for compiling Tcl. Less ↵jan.nijtmans2020-04-104-13434/+4989
| | | | | | | | | | use of ".." in the makefiles. Adapt all makefiles (unix/win) such that they can handle spaces in LIB_INSTALL_DIR/SCRIPT_INSTALL_DIR/MODULE_INSTALL_DIR, just by adding double-quotes in appropriate places.
* | Make sure that "Windows NT" is the only supported Windows platform. On ↵jan.nijtmans2020-04-031-9/+2
| | | | | | | | Windows CE, no chance that Tcl 8.6 will ever run .... Windows XP is the minimum now.
* | fixes [f583715154] - tclUnixSock.c: introduced ThreadActionProc considering ↵sebres2020-03-311-1/+48
| | | | | | | | a transfer of async-connecting channel between threads; cherry-picked some prevention against invalid thread owner addressed in [815e246806]
* | Remove MODULE_SCOPE function TclpSetInterfaces(), which is just empty. jan.nijtmans2020-03-312-207/+202
| | | | | | Increase size of MSG structure in Cygwin Notifier: Should conform with Win32 size. Some related code cleanup.
* | tcltest.tcl: Only adapt "filesMade" variable if the file to be removed is ↵jan.nijtmans2020-03-301-2/+2
| | | | | | | | | | actually found. tcltest 2.5.2 -> 2.5.3
* | Looks like this little hack is no longer necessary on current 32-bit Cygwin.jan.nijtmans2020-03-271-7/+0
| |
* | Fix harmless gcc warning, when compiling on Linuxjan.nijtmans2020-03-271-4/+4
| |
* | More uppercase HEX representations in source-code.jan.nijtmans2020-03-181-4/+4
| |
* | Merge 8.5. Mainly add Unicode 13 new plane 3 characters to the internal ↵jan.nijtmans2020-03-173-8/+8
|\ \ | |/ | | | | Unicode tables. Not used in 8.6 yet.
| * Use more uppercase hex (e.g. 0xFF) in stead of lowercase (0xff) consistantly.jan.nijtmans2020-03-093-8/+8
| |
* | Move setting of interpreter error-message from tclUnixSock.c to tclIO.c, ↵jan.nijtmans2020-03-051-31/+30
| | | | | | | | | | since the same should be done for all channel types. (Thanks, Don, for noticing this!) Also, ENOTCONN for a half-close should not be fatal, just a EINVAL, when doing it as part of a full close. See: [65c9cd1534]
* | The close routines of the "tcp" channel type were failing to record errordgp2020-03-041-1/+10
| | | | | | messages and error codes.
* | Implement WideSeekProc() for all channels which have a SeekProc(). Implement ↵jan.nijtmans2020-03-041-2/+15
| | | | | | | | | | Close2Proc() for all channels, as minimal wrapper around CloseProc(). Backported (with comments) and adapted from core-8-branch.
* | Merge 8.5jan.nijtmans2020-02-242-2/+2
|\ \ | |/
| * Add -Wpointer-arith warning to C-compile. Fix gcc(-9) warning.jan.nijtmans2020-02-242-4/+3
| |
* | Clean up some usage of DBGX: It isn't used any more on UNIXjan.nijtmans2020-02-042-26/+2
| |
* | Reset WSAGetLastError()/errno always, even when this error is not reported ↵jan.nijtmans2020-01-301-5/+6
| | | | | | | | due to the earlier error.
* | According to the ↵jan.nijtmans2020-01-301-17/+6
| | | | | | | | [https://core.tcl-lang.org/tcl/artifact?udc=1&ln=469-471&name=5ac7827cd282bbda|documentation], close2Proc(...., 0) should operate the same as closeProc(). Fix the UNIX/Windows socket channels to behave like that.