Activity for Scintilla

  • Vicente committed [eaa6c7]

    For rectangular selections, pressing Home or End now moves the caret to the Home

  • Neil committed [196655]

    Make argument name in declaration match definition.

  • Neil committed [b9e278]

    SciTE changelog.

  • Yury Dubinsky committed [d22d99]

    Bug [#1715]. Remember entered values in lists in more cases for find, replace

  • Neil Hodgson Neil Hodgson modified ticket #294

    Save find/replace history

  • Neil Hodgson Neil Hodgson posted a comment on ticket #294

    Option "save.find" implemented in 3.1.0.

  • Neil Hodgson Neil Hodgson modified ticket #1209

    [SciTE][windows]Support to execute python file in output window

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1209

    Committed change as [47bbc6].

  • Liang Bai committed [47bbc6]

    Feature [feature-requests:#1209]. Execute Python scripts directly by name when

  • Neil committed [c25d3c]

    SciTE change log.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1993

    For 01-cp-rm-via-SHELL.patch, its not good that a complex check is repeated before every COPY invocation. The command shells have different rules for things like quoting and brace expansion. The following produce different results with cmd and bash: echo "" echo \\ echo a{b,c} I don't know how safe these are on older Unix shells but something like this appears possible: ifeq (\,$(shell echo \\)) BASH = 1 endif

  • Tobias Kühne Tobias Kühne posted a comment on ticket #1993

    I was thinking mingw make brought the Utilities in Makefiles in, which proved wrong. Attached please find two alternative patches, one which determines COPY/DEL via the $(SHELL) variable (as already done for DEL in scite), and uses $(SHELL) in the comparison where copying, the other messing around with uname, which however is quite complicated, but for your information.

  • Neil Hodgson Neil Hodgson modified ticket #1998

    4.0.3 does not compile (MSVC 2017.5)

  • skunkos skunkos posted a comment on ticket #1998

    Indeed, you are correct. Can close.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1998

    std::clamp is C++17. Works due to -std:c++latest.

  • skunkos skunkos created ticket #1998

    4.0.3 does not compile (MSVC 2017.5)

  • Neil Hodgson Neil Hodgson modified ticket #1199

    Vertical line/column markers, or transparent background?

  • Neil Hodgson Neil Hodgson modified ticket #1200

    Contrast between sidebars (line and folding) is low

  • Neil Hodgson Neil Hodgson modified ticket #1203

    Add GETLENGTHNOSPACE

  • Neil Hodgson Neil Hodgson modified ticket #1206

    update cobol.properties (updated file attached)

  • Neil Hodgson Neil Hodgson modified ticket #1962

    SciTE cannot access c:\windows\system32

  • Neil Hodgson Neil Hodgson modified ticket #1970

    Inconsistent return value

  • Neil Hodgson Neil Hodgson modified ticket #1985

    Matlab code fold on while keyword

  • Neil Hodgson Neil Hodgson modified ticket #1208

    [Patch] Setting bookmark.symbol from properties file

  • Neil Hodgson Neil Hodgson modified ticket #1992

    SciTE: Stripping trailing spaces moves the caret.

  • Neil Hodgson Neil Hodgson modified ticket #1991

    SciTE displays incorrect column number if the caret is in virtual space.

  • Neil Hodgson Neil Hodgson modified ticket #1990

    scite: LDLIBS in makefile

  • Neil Hodgson Neil Hodgson modified ticket #1989

    SciTE.desktop is executable

  • Neil Hodgson Neil Hodgson modified ticket #1981

    Autocompletion list box still visible on macOS 10.13

  • Neil Hodgson Neil Hodgson modified ticket #1983

    Double tap a word is not working (Notepad++)

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1997

    Related bugs include [#1624] and [#1162].

  • Neil Hodgson Neil Hodgson modified ticket #1997

    Syntaxis error in Batch scripts

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1997

    Its helpful to include the text of an example as transcribing from an image takes time and may be inaccurate. The batch lexer is too simple for these sorts of expressions. It has no knowledge of how '(' can be used in batch files. Its possible that someone will come forward to work on these sorts of issues but I won't be working on this myself as I do not write complex batch files.

  • Juan Mier Juan Mier created ticket #1997

    Syntaxis error in Batch scripts

  • Scintilla released /SciTE/4.0.3/scite403.zip

  • Scintilla released /SciTE/4.0.3/scite403.tgz

  • Scintilla released /SciTE/4.0.3/wscite403.zip

  • Scintilla released /SciTE/4.0.3/gscite403.tgz

  • Scintilla released /scintilla/4.0.3/scintilla403.tgz

  • Scintilla released /scintilla/4.0.3/scintilla403.zip

  • Neil committed [9dfd91]

    Added tag rel-4-0-3 for changeset b5e1c1ee5b9d

  • Neil committed [71a51a]

    Added tag rel-4-0-3 for changeset 1bf8b7c50995

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1996

    Checked the Cocoa test application and it leads to an initial down arrow moving to the right by the width of the margins. Because the patch is inside the Editor contructor, it doesn't know the margins are in a separate view. This code should either be run later after marginInside is set by the most derived constructor or the constructor chain take arguments so that it is accurate by the time the Editor constructor is run. Another possibility would be to make lastXChosen be relative to the text origin...

  • Neil Hodgson committed [d5eb40]

    Eliminate calls of Scintilla APIs that have been removed.

  • Neil Hodgson committed [1bf8b7]

    Update Cocoa test app to Xcode 9.2 settings.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1993

    This makefile was developed for MinGW GCC with the native Windows shell CMD.EXE where there is no 'rm' or 'cp' available. Original form with bare "del /q" and "copy" Other people then wanted to use it with MSYS or on Linux so patches were accepted over time.

  • Van de Bugger Van de Bugger posted a comment on ticket #1996

    Does a low value for lastXChosen affect any behaviour? It probably does not affects stock Scintilla. It should just result in vertical key movement hugging the left side just as it would with an accurate value. I am experimenting with "wobble-free" vertical caret movement (it highly depends on lastXChosen) and noticed that lastXChosen is not correct at the very beginning. I thought is is better to make it correct from the beginning rather than check "is this correct?" every time. This looks wrong...

  • Tobias Kühne Tobias Kühne modified a comment on ticket #1993

    Maybe I am overlooking something, but are there any other systems this makefile is for? There is a different file for nmake, so there is only cygwin and MinGW to build in Windows as I understand it... So in which situations do del/xcopy have to be used at all? For the latter systems we would not need path translation at all; if rm/cp are missing there, I would consider this to be a misconfiguration. If so, the attached patch would work (with both make). Otherwise at least COPY should be resolved...

  • Tobias Kühne Tobias Kühne posted a comment on ticket #1993

    Maybe I am overlooking something, but are there any other systems this makefile is for? There is a different file for nmake, so there is only cygwin and MinGW to build in Windows as I understand it... So in which situations do del/xcopy have to be used at all? For the latter systems we would not need path translation at all; if rm/cp are missing there, I would consider this to be a misconfiguration. If so, the attached patch would work (with both make).

  • Liang Bai Liang Bai posted a comment on ticket #1209

    Thanks for your reply. I checked and found that I missed something in the description. Here is the steps: 1. Add D:\usr\bin to PATH environment variable. 2. Put pe++.py under D:\usr\bin. 3. Copy pe++.py to D:\, the new name is pe++a.py. 4. Open D:\test.c. 5. In the output window: you can run pe++a.py successfully while railed to run pe++.py.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1996

    Does a low value for lastXChosen affect any behaviour? It should just result in vertical key movement hugging the left side just as it would with an accurate value.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1996

    This looks wrong for margin outside platforms like Cocoa where the margins are in a separate view. The Editor constructor is run before ViewStyle::marginInside is reset. Maybe change constructors so that marginInside is initialized to correct value instead of being changed later in ScintillaCocoa constructor. ViewStyle::CalculateMarginWidthAndMask has already added up the widths of the margins to get fixedColumnWidth at the point the patch is setting lastXChosen.

  • Neil Hodgson Neil Hodgson modified ticket #1209

    [SciTE][windows]Support to execute python file in output window

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1209

    I'm not seeing this on Windows 10 1709 Fall Creators Update with Python 3.6.1 and current repository SciTE or SciTE 4.0.2. It does initially return ERROR_FILE_NOT_FOUND so gets rerun with cmd/c. With this script saved into pe++.py: import sys print("a", sys.argv) Running it by typing "pe++.py test.cpp" gives these results: pe++.py test.cpp >pe++.py test.cpp a ['C:\\u\\hg\\pe++.py', 'test.cpp'] >Exit code: 0 Its possible that the .py extension is not associated with Python.exe on your machine maybe...

  • Van de Bugger Van de Bugger created ticket #1996

    Scintilla: Editor::lastXChosen has incorrect initilal value

  • Neil committed [b5e1c1]

    Updates for 4.0.3.

  • Neil committed [7295a8]

    Updating for 4.0.3.

  • Neil Hodgson Neil Hodgson modified ticket #1995

    SelectionRange::Length() gives false results

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1995

    Before running .appveyor-prepare-win.ps1, the Textilosaurus source code must have been installed into C:\textilosaurus. Otherwise .\resources\scripts\7za\7za.exe will not be found and errors out. At 380MB, the archive is too large for me to download yet again just to show accurate error messages.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1995

    Built SciTE with /O1 + /GL and the issue does not occur.

  • Neil Hodgson Neil Hodgson modified ticket #1983

    Double tap a word is not working (Notepad++)

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1983

    Committed above fix (with some extra brackets) as [1280ef].

  • Neil Hodgson Neil Hodgson modified ticket #1983

    Double tap a word is not working (Notepad++)

  • Dimitar Radev committed [1280ef]

    [Bug #1983] Fix double tap word selection on Windows 10 1709 Fall Creators Update.

  • Dimitar Radev Dimitar Radev modified a comment on ticket #1983

    I think this should be left as word select as this is what i intend to do when double tapping. Unless maybe the word selection is made to happen on buttonUp instead of the second button down, such that down + big move without up is drag select. By the way on my laptop (HP Probook with sinaptics toucpad) now doing double click , but not releasing the touchpad doesnt do word select but drag select, unlike the same action with the mouse. To me it looks like this tap twice to multi select does this -...

  • Dimitar Radev Dimitar Radev modified a comment on ticket #1983

    I think this should be left as word select as this is what i intend to do when double tapping. Unless maybe the word selection is made to happen on buttonUp instead of the second button down, such that down + big move without up is drag select. By the way on my laptop (HP Probook with sinaptics toucpad) now doing double click , but not releasing the touchpad doesnt do word select but drag select, unlike the same action with the mouse. To me it looks like this tap twice to multi select does this -...

  • Dimitar Radev Dimitar Radev posted a comment on ticket #1983

    I think this should be left as word select as this is what i intend to do when double tapping. Unless maybe the word selection is made to happen on buttonUp instead of the second button down, such that down + big move without up is range select. By the way on my laptop (HP Probook with sinaptics toucpad) now doing double click , but not releasing the touchpad doesnt do word select but drag select, unlike the same action with the mouse. To me it looks like this tap twice to multi select does this...

  • skunkos skunkos posted a comment on ticket #1995

    I do not know, but there is one.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1995

    Note that this is not likely bug in compiler, but rather symptom of a bug in your code. What is the actual bug then?

  • skunkos skunkos modified a comment on ticket #1995

    OK, I narrowed it down. The problem is definitely triggered by combination of /O1 AND /GL compiler flags. When built without /O1 + /GL, everything works, when build with /O1 + /GL bug is there. Note that this is not likely bug in compiler, but rather symptom of a bug in your code. It would be quite a miracle to discover the actual bug in the compiler.

  • skunkos skunkos posted a comment on ticket #1995

    OK, I narrowed it down. The problem is definitely triggered by combination of /O1 AND /GL compiler flags. When built without /O1, everything works, when build with /O1 bug is there. Note that this is not likely bug in compiler, but rather symptom of a bug in your code. It would be quite a miracle to discover the actual bug in the compiler.

  • skunkos skunkos posted a comment on ticket #1995

    Let me perform some more testing.

  • skunkos skunkos posted a comment on ticket #1995

    What exact error did you get?? There are literally not C: assumptions, all code compiles fine on AppVeyor.

  • skunkos skunkos posted a comment on ticket #1995

    I would bet that compiler in LTCG/GL mode will perform some kind of optimizations (branching prediction maybe) which will actually produce bad result. Or maybe the comparison operator < implementation is not entirely correct. Anyway, what could be possibly wrong with using proposed fix??? It should produce mathematically correct results in all hyphotetical situations, while actually being a bit faster than your implementation.

  • skunkos skunkos posted a comment on ticket #1995

    No way to debug for me, it happens only on release/LTCG configurations. I does not seem to be linker-specific because it happens even when compiled on completely different machine. Even compilling totally minimal QT/scintilla application with LTCG and sources directly compiled in reproduces the issue for me.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1995

    Can you see any good reason for SelectionRange::Length to return 0? Because I can't. That means this is likely failing for a bad reason, such as a compiler failure, link failure, or memory bug. Since the failure hasn't been located, replacing the implementation of SelectionRange::Length is likely to just move the problem to somewhere we don't see it for now. Running the debugger over the call location and switching to assembler may give more information on what is occurring. Try turning off /GL and...

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1995

    Tried to build and the scripts and instructions are incomplete or make bad assumptions like the current disk being C:. Build scripts should be complete or document their requirements. Gave up.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1983

    The trace shows an WM_LBUTTONDOWN with a timestamp before the previously received WM_LBUTTONUP. A subsequent WM_MOUSEMOVE is timestamped more than 2 seconds earlier. The "tap twice and drag to multi-select" option appears to be doing something complex here and I haven't been able to decipher it. In this mode, perhaps Scintilla's response to this event sequence should not be to select a word but, instead, to start a range (multi-select) selection. Casting from an unsigned expression as suggested appears...

  • Liang Bai Liang Bai created ticket #1209

    [SciTE][windows]Support to execute python file in output window

  • Kein-Hong Man Kein-Hong Man posted a comment on ticket #1994

    To anyone interested in implementing this: Some other systems (this SF ticket's markdown system, for example) appear to have taken the "mixing strings and statement blocks" approach, that is, command strings are highlighted like normal code. I guess this is akin to code and data mixing in HTML/JavaScript. So in the last post, the two pieces of the one double-quoted string entity are separated by normally highlighted code (probably). Sometimes it's rather unclear where a string starts and where it...

  • Dimitar Radev Dimitar Radev posted a comment on ticket #1983

    http://prntscr.com/ia5egq Looks like the issue is related to some crazy timing information returned by GetMessageTime so time between last message and current message gets negative and wraps around as the comparison to doubleclick time is an unsigned one.. These casts for signed comparison seems to fix it src/Editor.cxx line 4458 if ((((int)(curTime - lastClickTime)) < (int)Platform::DoubleClickTime()) && Close(pt, lastClick, doubleClickCloseThreshold)) {

  • skunkos skunkos posted a comment on ticket #1995

    Let me know if you need any extra information. It is 100% reproducible, simple fix: Sci::Position SelectionRange::Length() const { return std::abs(anchor.Position() - caret.Position()); }

  • Kein-Hong Man Kein-Hong Man posted a comment on ticket #1994

    In the following example, we can mix command strings and double-quoted strings, and comments are recognized within the command strings: 1 2 3 4 5 6 7 8 9 10 11 12 13 14#!/bin/bash a=$( # comment level 1 echo 'level1' b="double_quote_start $( # comment level 2 echo 'level2' ) double_quote_end" echo $b ) echo $a echo "end of test" This illustrates the problem with the bash method of parsing strings. What rules do the double-quoted string assigned to b follow? It seems that bash switches string rules...

  • Kein-Hong Man Kein-Hong Man posted a comment on ticket #1994

    Here is a tested example of nested $() and backtick strings: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23#!/bin/bash a=$( # comment level 1 echo 'level1' echo $( # comment level 2 echo 'level2' ) echo ` # comment level 2b echo 'level2b' echo ` # comment level 3 echo 'level3' echo $( # comment level 4 echo 'level4' ) ` ` ) echo $a echo "end of test" The comments are all recognized as valid comments by bash. As such, inside such comment lines, string delimiter characters are not semantically...

  • Kein-Hong Man Kein-Hong Man posted a comment on ticket #1994

    I've looked over my notes and test cases and tried out some "what else" cases that bugged me. This may (or may not) help anyone who wishes to attack this issue. First, the bash docs mentions $(COMMAND) when I worked on this feature area. It may have changed since then. I saw $(COMMAND) and expected coders to write stuff of limited complexity inside the $(). But now it appears bash just adjusts its parsing state. Things can be very complex inside the $(), it basically allows arbitrary code. Did that...

  • skunkos skunkos modified a comment on ticket #1995

    Steps to reproduce: Install custom Qt 5.10 LTCG build - https://bitbucket.org/skunkos/qt5-minimalistic-builds/downloads/qt-5.10-static-ltcg-msvc2017-x86_64.7z Build any sample app with scintilla-qt and this Qt version, for example Textilosaurus with these steps: https://github.com/martinrotter/textilosaurus/blob/2018.01.31/resources/scripts/.appveyor-prepare-win.ps1 https://github.com/martinrotter/textilosaurus/blob/2018.01.31/resources/scripts/.appveyor-build-win.ps1 Bug is really there. I used...

  • skunkos skunkos posted a comment on ticket #1995

    Steps to reproduce: Install custom Qt 5.10 LTCG build - https://bitbucket.org/skunkos/qt5-minimalistic-builds/downloads/qt-5.10-static-ltcg-msvc2017-x86_64.7z Build any sample app with scintilla-qt and this Qt version, for example Textilosaurus with these steps: https://github.com/martinrotter/textilosaurus/blob/2018.01.31/resources/scripts/.appveyor-prepare-win.ps1 https://github.com/martinrotter/textilosaurus/blob/2018.01.31/resources/scripts/.appveyor-build-win.ps1 Bug is really there. I used...

  • skunkos skunkos posted a comment on ticket #1995

    I used bundled qt, latest stable. Build with static qt with ltcg. On 2 Feb 2018 23:21, "Neil Hodgson" nyamatongwe@users.sourceforge.net wrote: Textilosaurus seems to be using Qt on 64-bit Windows. I tested the Haven demonstration program with Qt 5.8.0 building with win32-msvc2015 spec 64-bits on Windows 10 with current Scintilla code from the repository and could not reproduce the ["aa"; Select All; "a"] result in either release or debug builds. I am using Scintilla's bundled Qt platform, not Riverbank's...

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1995

    Textilosaurus seems to be using Qt on 64-bit Windows. I tested the Haven demonstration program with Qt 5.8.0 building with win32-msvc2015 spec 64-bits on Windows 10 with current Scintilla code from the repository and could not reproduce the ["aa"; Select All; "a"] result in either release or debug builds. I am using Scintilla's bundled Qt platform, not Riverbank's QScintilla.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1993

    If minw32-make works fine then I'm not seeing a strong reason to change. If there is a reason to change, then this should be simplified and documented a bit more. Possibly detangle the OS test (which is needed for grabbing Windows version to determine theming) from the shell (either CMD or bash (or similar)) which determines DEL and COPY and whether path translation is needed.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1972

    I think that was because debug builds don't use /GL and the linker knows what to do when it enounters modules compiled with /GL. If anyone wants to write a nice patch that applies /LTCG just for release builds and leaves everything else as-is then that can be looked at after the next release.

  • Neil Hodgson Neil Hodgson posted a comment on ticket #1995

    This doesn't reproduce for me on macOS 10.13 with Xcode 9.2 or on Windows 10 with MSVC 2017.5 in 64 or 32 bits, debug or release. On which platform and compiler are you seeing this behaviour?

  • Tobias Kühne Tobias Kühne posted a comment on ticket #1993

    Sorry, maybe I should have had a closer look at the patch before submitting. I already made it when 3.6.2 was current so I have to try hard to remember what it exactly was about. Maybe the problem is the invocation of make. I guess at the time I created the patch I did not know about mingw32-make, which has to be used instead of make - which in my case promptly found the cygwin make (D'OH!!). This leads at least to xcopy to be complaining about 'invalid number of parameters' - probably because of...

  • skunkos skunkos posted a comment on ticket #1972

    If you compiled some dependent libraries with /GL switch then you need to link against those libraries with /LTCG. If you do not, then MSVC linker will do it for you, but displays this warning.

  • Kein-Hong Man Kein-Hong Man posted a comment on ticket #1994

    Thank you for the bug report. This is due to the lexer historically trying to analyze command strings as a kind of string, and not something that needed full-blown nested parsing. It detects the single quote as the start of a string, because in bash there are cases where different types of strings can be nested and their escape behaviour changes accordingly. Unfortunately, now it appears that I need to scan for # comments also in command strings, and god knows what else. Not easy to do without doing...

  • skunkos skunkos posted a comment on ticket #1995

    Btw, note that in https://pastebin.com/uKVWhzsB there are also occurrences of Length() replaced by my custom code which fixes the issue

  • skunkos skunkos created ticket #1995

    SelectionRange::Length() gives false results

  • Erich Eckner Erich Eckner created ticket #1994

    comment issue in syntax highlighting of shell/bash

  • Neil committed [431b81]

    Implement SC_DOCUMENTOPTION_STYLES_NONE.

  • Neil committed [3c03ec]

    Scintilla API update.

1 >