![]() |
#91 |
Wannabe Connoisseur
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 426
Karma: 2516674
Join Date: Apr 2011
Location: Geelong, Australia
Device: Kobo Libra 2, Kobo Aura 2, Sony PRS-T1, Sony PRS-350, Palm TX
|
I have seen this with sorta relatively quick finds without holding down a key combo. See my description of ctrl-] and tap,tap,tap above. The press-and-hold ctrl-g just happens to make it relatively reproducable instead of happens some times but not sure how.
|
![]() |
![]() |
![]() |
#92 | |
Connoisseur
![]() Posts: 84
Karma: 10
Join Date: Jul 2012
Location: Florida USA
Device: Android, kindle
|
Quote:
But other than that all of the other things that I use that were messed up have been corrected. Except for a very minor one. On earlier versions of Sigil when in spellcheck and used the 'Change Selected Word To' button. The next misspelled word that was shown was also highlighted in CV. Like I said minor. Thank You Very Much. Paul |
|
![]() |
![]() |
Advert | |
|
![]() |
#93 | |
Sigil Developer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 8,789
Karma: 6000000
Join Date: Nov 2009
Device: many
|
No that User Preference is always used in CodeView if set. The colour changes are due to active vs inactive palette colours. When using Spellcheck or Saved Searches, those windows always have focus and so CodeView is inactive as it uses inactive palette colours to indicate that.
As for selecting the next word, that does happen but only when the "All Words" checkbox is not checked. If only showing misspelled words, once that word is corrected it is removed from the list advancing the selection. When the "all words" is selected, fixing a misspelled word does not advance anything. In addition, the search for the selected word automatically with a single click is gone and has been fow a few releases now as it tremendously slowed down spell checking when huge word lists (huge epubs) are being used (the search for the next instance in the text took too long making hopping around too slow). That is now expected behaviour. Double-clicking now forces the search to be done if you really need to see context. This hopping around delay is pretty much unnoticeable in short epubs/word lists butwas intolerable in very large epubs, thus the change. Thanks for testing and confirming your issue was fixed. These fixes will appear in the upcoming Sigil-1.9.2 follow-on release that will hopefully be out soon. Quote:
Last edited by KevinH; 03-09-2022 at 12:15 PM. |
|
![]() |
![]() |
![]() |
#94 |
Connoisseur
![]() Posts: 74
Karma: 10
Join Date: Nov 2012
Device: none
|
Thanks Kevin! I've always used switches to indicate placeholders, regardless of whether I fill them using Regex or by other means. Yes, I agree with you and DiapDealer that this a sloppy habit, but prior versions of Sigil covered for me, so thank you for that too. To tell the truth, I was more concerned about how newline (\n) was executed, or not, and reported what I saw so as to contribute to background knowledge.
As always, thank you Team Sigil for the care shown in meeting user concerns, no matter how small. |
![]() |
![]() |
![]() |
#95 | |
Sigil Developer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 8,789
Karma: 6000000
Join Date: Nov 2009
Device: many
|
I am glad you can recreate it. If it does not happen with Linux and macOS, then it may be a Qt issue.
A crash dump file and backtrace from it would surely help! Does this thing still crash when using a Qt6 build of current master on Windows? If not, maybe we can track down the Qt fix that prevents that segfault by searching Qt 6 windows specific code for changes since Qt 5.12.9. Other than that we are grasping at straws. I have no idea what the equivalent of lldb (mac) and gdb (linux) are on Windows anymore. They must have something equivalent. Quote:
|
|
![]() |
![]() |
Advert | |
|
![]() |
#96 | |
Sigil Developer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 8,789
Karma: 6000000
Join Date: Nov 2009
Device: many
|
Thanks for your report. It matches what DiapDealer is seeing.
From a web search of your exception code, it is a stack overflow deep inside QtGui: ExceptionCode: c00000fd (Stack overflow) So my guess is Qt is storing the repeated key sequence someplace and search slows the processing down enough that it fills the stack frame where the key sequences are being accumulated and eventually causes a stack over flow. Either that or there is a recursive call being invoked in Qt that never gets to unwind before the call stack overflows the process stack size. Not sure what we can do about it, as the issue is deep inside Qt Windows specific code. Delaying after a key is processed may actually cause more build ups and faster stack overflow in Qt. On Windows, is there any way to set the process or thread stack size allocation? Kevin Quote:
Last edited by KevinH; 03-09-2022 at 01:02 PM. |
|
![]() |
![]() |
![]() |
#97 |
Sigil Developer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 8,789
Karma: 6000000
Join Date: Nov 2009
Device: many
|
I think for Windows, when the ctrl-g or ctrl-shift-g shortcuts are received in MainWindow before passing them onto search, we may want to send them to a routine that does nothing and returns quickly, and then pass only new ones on to search once the previous search is complete.
That should prevent any overflow of queued event signals waiting to be delivered and hopefully prevent the stack overflow from happening. For this to work, MainWindow just needs to know if search is running or not to know how to treat it when it gets the shortcut. I see a quick way to test this hypothesis. I will push those changes and hopefully people can check to make sure the crashing stops on Windows. I am building a new test build now to see if it helps prevent the crashing, I will post the link when ready. Last edited by KevinH; 03-09-2022 at 02:24 PM. |
![]() |
![]() |
![]() |
#98 |
Sigil Developer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 8,789
Karma: 6000000
Join Date: Nov 2009
Device: many
|
Hopefully Final Test Build
Hi All,
As far as my testing shows, this one seems to fix all of the bugs reported so far. This includes a fix for capture groups \1 being improperly processed when no capture groups in regex. All of the debug output has been removed so it may react a bit quicker. And it has a potential "fix" for the ctrl-g (held down) segfault for Windows but whether this really fixes the problem seen remains to be tested. The test builds can be downloaded from my repo here: https://github.com/kevinhendricks/Bu...OnMac/releases - Sigil-master-f2157d4-build_num-378-Mac.tar.xz - Sigil-master-f2157d4-build_num-378-Win-Setup.exe Again thanks to everyone who tests and helps make Sigil better! If you have a Windows box and could create the ctrl-g repeated too quickly crash, please give this test version a try and let me know if it prevents the crash or if the crash still happens just someplace else. Last edited by KevinH; 03-09-2022 at 03:29 PM. |
![]() |
![]() |
![]() |
#99 |
Guru
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 842
Karma: 3335974
Join Date: Jan 2017
Location: Poland
Device: Various
|
I confirm – no crash after long press Ctrl-G!
Edit: I also did a private build and it works too! |
![]() |
![]() |
![]() |
#100 |
Sigil Developer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 8,789
Karma: 6000000
Join Date: Nov 2009
Device: many
|
|
![]() |
![]() |
![]() |
#101 |
Grand Sorcerer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 28,587
Karma: 204624552
Join Date: Jan 2010
Device: Nexus 7, Kindle Fire HD
|
I've gotten no long-press crtl-g crashes with the latest test version, but this is a different Windows machine from the one where I was able to reproduce the crash consistently.
|
![]() |
![]() |
![]() |
#102 |
Sigil Developer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 8,789
Karma: 6000000
Join Date: Nov 2009
Device: many
|
The change just drops shortcuts that are delivered when the previous shortcut request had not completed yet. I applied it for all platforms since it can not hurt. If the screen has not updated from the last search yet, another search makes no sense.
Hopefully this will fix it on your other machine as well, it can not hurt! |
![]() |
![]() |
![]() |
#103 | |
Connoisseur
![]() Posts: 84
Karma: 10
Join Date: Jul 2012
Location: Florida USA
Device: Android, kindle
|
Quote:
However, I went back to a previous version (1.3) and did spellcheck and saved searches and both of them used the highlight color displayed in the CV because that version put the CV in focus and didn't keep focus themselves. Paul |
|
![]() |
![]() |
![]() |
#104 | |
Connoisseur
![]() Posts: 84
Karma: 10
Join Date: Jul 2012
Location: Florida USA
Device: Android, kindle
|
Quote:
Find & Replace Spellcheck and Saved Searches all kept focus. But they displayed the found item in the CV using the line highlight color. Spellcheck highlights the line and the incorrect spelled word on a "single click". Paul |
|
![]() |
![]() |
![]() |
#105 | |
Sigil Developer
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() Posts: 8,789
Karma: 6000000
Join Date: Nov 2009
Device: many
|
Just to be clear.
The different line highlight in CodeView due to focus came in Sigil-1.5 to make focus difference visually clearer. And the change to need double-click to sync Spellcheck word in CodeView came in Sigil-1.7 to make hopping around and paging very long lists of much words faster. So both are changes not bugs (minor or otherwise) and they work as expected and have been that way for a couple of releases over the past year or more. Kevin Quote:
|
|
![]() |
![]() |
![]() |
|
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
Sigil-1.9.0 Release | DiapDealer | Sigil | 101 | 03-11-2022 07:20 PM |
Sigil-1.8.0 Release | DiapDealer | Sigil | 59 | 12-24-2021 10:38 PM |
Sigil-0.9.0 Release | KevinH | Sigil | 76 | 11-14-2015 06:45 PM |
New Sigil Release: Sigil-0.8.901 | KevinH | Sigil | 137 | 11-03-2015 09:42 AM |
Sigil 0.2.0 Release Candidates | Valloric | Sigil | 46 | 05-10-2010 09:55 PM |