Sigil running multiple instances

[BetterRed]

Originally Posted by DiapDealer

I see a lot of scattering of webengine cache files on both Linux and Windows. I've always meant to try Kovid's approach at unifying on Windows (requires patching qtbase), but I've never gotten around to it.

...

I get a calibre-ebook.com folder in C:|Users\<??????>\AppData\Roaming despite the fact that I relocate calibre's preferences and cache via its EnvVar directives, if I delete the folder it gets recreated when I use the calibre Viewer or Editor… even if I use calibre's supposedly self-contained portable edition.

Sigil's the same, even though I relocate it's Preferences from C:\Users\<??????>\AppData\Local via its EnvVar directive, I still get a sigil-ebook folder in C:\Users\<??????>\AppData\Roaming.

FTR: I have no problems running multiple instances of Sigil 2.4.2… including via calibre's built-in Open with.

BR

[DiapDealer]

@KevinH: I may have led you astray a bit. I did a full rebuild (without the latest off-the-record change) and I'm no longer getting the "deadline expired" message. It's still not deleting the empty 'Preview- +uuid' folders from C:/Users/<USER>/AppData/Roaming/sigil-ebook/sigil/QtWebEngine after Sigil closes, but the infinite loop does NOT seem to be happening with your previous change.

Sorry for any potential confusion. Should I still try the new off-the-record approach or do you want to revert and do more testing?

I think the new QWebEngineProfileBuilder class in 6.9 is intended to fix this exact sort of issue:

https://codereview.qt-project.org/c/...ngine/+/597808

[DiapDealer]

Originally Posted by BetterRed

FTR: I have no problems running multiple instances of Sigil 2.4.2… including via calibre's built-in Open with.

The problem doesn't manifest until after the official 2.4.2 release. If we can figure it out, it won't manifest in ANY official release.

[KevinH]

Please try the off the record approach. If that works we at least have a safety net.

We can then revert and try to track down exactly where the damn cache files are going on Windows so that they are properly cleared.

According to latest qtbase code it should go here:

extracted from qtbase source in github from src/corelib/io/qstandardpaths_win.cpp

QString QStandardPaths::writableLocation(StandardLocation type)
{
    QString result;
    switch (type) {
    case CacheLocation:
        // Although Microsoft has a Cache key it is a pointer to IE's cache, not a cache
        // location for everyone.  Most applications seem to be using a
        // cache directory located in their AppData directory
        result = sHGetKnownFolderPath(writableSpecialFolderId(AppLocalDataLocation));
        if (!result.isEmpty()) {
            appendTestMode(result);
            appendOrganizationAndApp(result);
            result += "/cache"_L1;
        }
        break;

But it somehow ends up in roaming and seemingly without the final "cache" folder?

Does this folder even exist?

C:\Users\<??????>\AppData\Roaming\sigil-ebook\sigil\cache\

[DiapDealer]

Originally Posted by KevinH

Does this folder even exist?

C:\Users\<??????>\AppData\Roaming\sigil-ebook\sigil\cache\

No. The cache folder does not currently exist in either of the locations that this cache stuff is getting written to.

In roaming, things are being written to C:\Users\<??????>\AppData\Roaming\sigil-ebook\sigil\QtWebEngine\

For the record: the 'cache' folder has been present in Sigil's regular prefs directory (C:\Users\<??????>\AppData\Local\sigil-ebook\sigil\) when we've been testing different strategies. But before your latest commit. It did not exist.

I'll try the latest commit (off-the record cache) and get back to you.

[DiapDealer]

OK. That last commit seems to be working. Multiple instances can be launched with no 'deadline expired' messages logged. The only thing (re)created in C:/Users/<USER>/AppData/Roaming is an empty 'sigil-ebook/sigil' folder

Preview-Cache is still being created in Sigil's prefs location (C:/Users/<USER>/AppData/Local/sigil-ebook/sigil), but I see from master that it's still being created if it doesn't exist. Nothing is being written to it. No 'cache' folder and no 'QtWebEngine' folder has been (re)created.

[KevinH]

Okay. So with some cleanups, we have a fallback safety net with off-the-record.

I have now hopefully pushed to master a fixed version of the previous attempt (named storage with unique names) and then we can decide which one to use later.

If you have time, please give it a try and let me know if this one works.

Otherwise I will study the code and try to figure out something else tomorrow.

Thanks!

[DiapDealer]

I think we might be closer. There are no infinite loops and multiple instances work.

The named storage is created in Sigils prefs directory (under the 'cache' folder) and they are cleaned up properly after Sigil exits.

The only problem is that the named storage is being duplicated in a C:\Users\<??????>\AppData\Roaming\sigil-ebook\sigil\QtWebEngine (not the proper Sigil prefs location) and they are NOT being cleaned up after Sigil exits. They are being orphaned. There's never any content in the orphaned folders. Even when Sigil is still open. It's like they're being created there before the code that properly creates the caches in Sigil's prefs directory takes effect or something. Very weird.

I think it might be the very act of trying to change these default locations that's introducing these anomalies on Windows.

We'll have to revisit tomorrow sometime.

EDIT: Sorry, I forgot to attach the debug output. It's reporting the correct paths (not the stuff in Roaming) by the way.

[KevinH]

Wow, I stopped trying to set it and instead just let it create itself. Then I ask the profile what cachePath it will be using and it reports the local app data instead of the roaming one, and it adds the "cache" as well. This only happens with Windows. It reports correctly on Linux and MacOS.

But we know this is not correct at all. So QtWebEngine for windows is saying its cache is one place but it is actually in another!

I will have to either hard code in the roaming bit or try to find where that path is being created somehow. This is a mess.

Tomorrow, I will try to track down where this is being set in the Windows code in QtWebEngine.

[DiapDealer]

Originally Posted by KevinH

Wow, I stopped trying to set it and instead just let it create itself. Then I ask the profile what cachePath it will be using and it reports the local app data instead of the roaming one, and it adds the "cache" as well. This only happens with Windows. It reports correctly on Linux and MacOS.

But we know this is not correct at all. So QtWebEngine for windows is saying its cache is one place but it is actually in another!

I will have to either hard code in the roaming bit or try to find where that path is being created simehow. This is a mess.

Tomorrow, I will try to track down where this is being set in the Windows code in QtWebEngine.

For the record, the location being reported in the debug log is the actual location (in Sigil's prefs location) where the caches are actually being created (under the cache prefix), used, and then cleaned up properly. The problem is that the directories are also being created (but unused), and orphaned in the Roaming location. The local app data location is the correct location.

[DNSB]

For what it may be worth, I am still seeing the Debug: View Preview Cache Clear failed - deadline expired on the code I downloaded from GitHub this evening.

[DiapDealer]

Are you doing a full rebuild, or are you using an existing build directory and doing a partial build with just the changed files? I was still seeing the deadline expired messages until I deleted everything in my build directory, did a new cmake configure and a full rebuild.

[DNSB]

I delete the contents of my sigil-src directory and then unzip the sigil-master directory in the downloaded .zip file into that directory. Run cmake/nmake/nmake makeinstaller and install the output file.

I also deleted the contents of the Program Files\Sigil directory and re-installed and still no joy.

[DiapDealer]

Well shoot. That indicates there might be a race condition that one of us is losing/winning.

[KevinH]

Originally Posted by DiapDealer

For the record, the location being reported in the debug log is the actual location (in Sigil's prefs location) where the caches are actually being created (under the cache prefix), used, and then cleaned up properly. The problem is that the directories are also being created (but unused), and orphaned in the Roaming location. The local app data location is the correct location.

Oh, then I misunderstood you. So I just need to clean up the Roaming location as an extra location. I can add that. I thought the actual internal cache files were being stored there.