Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix editor sync bug, when model file was edited outside of eclipse. #3108

Conversation

mehmet-karaman
Copy link
Contributor

This bugfix contains the fix for unsync workspace files wrt. the file system (see the
DirtyStateEditorSupportIntegrationTest.testModifyFileInExternEditor) and a one liner for beeing able to save the file when the changes were ignored. (see the ignored test case
DirtyStateEditorSupportIntegrationTest.testModifyDirtyFileInExternEditor).

The issue for this fix can be found here:
#2385

@cdietrich cdietrich requested a review from szarnekow July 23, 2024 11:08
synchronizationStamp = ((URIInfo) info).synchronizationStamp;
} else {
return super.isSynchronized(element);
if (element instanceof IFileEditorInput) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I personally avoid many nested if levels, so I would revert the condition and also use modern instanceof
if (!(element instanceof IFileEditorInput input)){
return super.isSynchronized(element);
}

// continue with the new code
// "input" is already defined here with the right type

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Currently the project uses java 11. Am I allowed to upgrade to java 17 or 21?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Am I allowed to upgrade to java 17 or 21?

I would say, definitely not in the context of this ticket, but it is a good question and probably deserves a dedicated ticket in Xtext.

Note that Eclipse Platform is switched to Java 17 in Eclipse 4.28, including ecj (see eclipse-jdt/eclipse.jdt.core#886).

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

there is already a ticket for drop 11 17 minimum
#2686

@cdietrich
Copy link
Contributor

i am only aware of #3102
the others are new to me

@mehmet-karaman mehmet-karaman force-pushed the fix_xtext_editors_sync_ext_editing branch from 52d207b to f34d2fe Compare July 23, 2024 13:38
@cdietrich
Copy link
Contributor

cdietrich commented Jul 23, 2024

started j11 build on main
https://ci.eclipse.org/xtext/job/xtext/job/main/1325/

i also wonder if the base branch should be rebase.

@cdietrich
Copy link
Contributor

cdietrich commented Jul 23, 2024

i can reproduce locally, but no idea why
@LorenzoBettini do you know why this
https://ci.eclipse.org/xtext/job/xtext/job/main/1325/consoleFull
seems to run with 17 and not 11?

maybe something is broken with that
why was it working here yesterday:
https://ci.eclipse.org/xtext/job/xtext/job/cd-require-mocha-bump/1/
(reran this one)

and also scheduled
https://ci.eclipse.org/xtext/job/xtext/job/PR-3108/ to run with 17 explicit and 2024-06

@LorenzoBettini
Copy link
Contributor

i can reproduce locally, but no idea why @LorenzoBettini do you know why this https://ci.eclipse.org/xtext/job/xtext/job/main/1325/consoleFull seems to run with 17 and not 11?

@cdietrich from what I see

09:52:11  [INFO] Command line:
09:52:11  	[/opt/tools/java/temurin/jdk-11/latest/bin/java

it runs using JDK 11 (of course, the build runs with Java 17, but the toolchain uses JDK11)

@mehmet-karaman
Copy link
Contributor Author

The failing tests are running successful on my branch. I am going to rebase on the latest state and try agian.

@mehmet-karaman mehmet-karaman force-pushed the fix_xtext_editors_sync_ext_editing branch from f34d2fe to 3f5e61d Compare July 24, 2024 08:06
@mehmet-karaman
Copy link
Contributor Author

Tests are also successful after rebase.. Pushed the branch again.

@iloveeclipse
Copy link
Contributor

Jenkins is fine, new test is executed. Whatever the pipeline is doing and why it is failing is not our beer, it seem to fail on all other recent PR's too.

Xtext committers: Could we proceed with the review?

@mehmet-karaman
Copy link
Contributor Author

@szarnekow @cdietrich Do i have to do something on my side? Does this PR has any problems?

@cdietrich
Copy link
Contributor

problem: ustream is never green cause all of the jdt changes.
i want sebastian to have a review cause i cannot judge if it may have bad side effect i dont see.

@cdietrich
Copy link
Contributor

@mehmet-karaman can you please rebase

@mehmet-karaman mehmet-karaman force-pushed the fix_xtext_editors_sync_ext_editing branch from 3f5e61d to d7e4fd9 Compare August 9, 2024 07:08
@iloveeclipse
Copy link
Contributor

All tests are green now. Cool.

@iloveeclipse
Copy link
Contributor

@cdietrich, @szarnekow : could we proceed with the review now?

@cdietrich
Copy link
Contributor

as i said i cannot judge it it will break something. and in my 12 and 13th working hour i also have no capacity to think about it.

and sebastian has no time too i assume

}

/**
* This test needs a manual button click.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This comment sounds strange: does a human have to click on the button to make the test proceed?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Exact, this is the reason the test is "ignored".
Unfortunately a confirmation dialog pops up during the test, so the test is there to demonstrate the fix.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Mine is not meant to be a review, but I don't understand what the test does: if it's ignored, it doesn't test anything.

I don't think clicking on a dialog button is (easily) doable in a plug-in test, but it's easy in a SWTBot test.

Maybe a SWTBot test to effectively verify the fix would be better.

But, again, I did not follow the whole bug and fix.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sure, therefore there is one automated test and one that needs manual interaction.
Not sure if SWTBot is used in Xtext code.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We use SWTBot tests, just a few, to test Xtend dialogs.

My main concern is that there's no automatic test for verifying the fix, again, if I understand correctly.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There is one test which is executed automatically, which tests the basic functionality.. The ignored test is testing the case when the editor is dirty and the file is out of sync case.. I didn't saw an example how to automate a click on a button, without introducing a new dependency.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The test is green without the changes in XtextEditor. Do you have an idea how to assert the necessity of the the overridden performSave?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@szarnekow : also the manual test is green without changes?

Copy link

@memin20 memin20 Sep 4, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For this the ignored test has to be unignored and executed manually..

I am Mehmet, Just writing on my phone because my doctors appointment..

@mehmet-karaman mehmet-karaman force-pushed the fix_xtext_editors_sync_ext_editing branch from d7e4fd9 to a73abd6 Compare August 21, 2024 09:55
Copy link

github-actions bot commented Aug 21, 2024

Test Results

  6 460 files  ± 0    6 460 suites  ±0   3h 1m 54s ⏱️ - 17m 21s
 43 242 tests + 2   42 658 ✅ +1    584 💤 +1   0 ❌ ±0 
170 114 runs  +17  167 740 ✅ +4  2 358 💤 +4  12 ❌ +5  4 🔥 +4 

Results for commit 5fa196f. ± Comparison against base commit 197f6cc.

♻️ This comment has been updated with latest results.

@mehmet-karaman
Copy link
Contributor Author

I am going to rebase against the latest master.

@mehmet-karaman mehmet-karaman force-pushed the fix_xtext_editors_sync_ext_editing branch from a73abd6 to be83d08 Compare August 21, 2024 11:38
@iloveeclipse
Copy link
Contributor

@szarnekow, @cdietrich : anything is missing here? Could we continue with review or merge?


// Check has to fit the following method:
// org.eclipse.ui.editors.text.FileDocumentProvider.checkSynchronizationState(long, IResource)
boolean isSynchronized = synchronizationStamp == getSynchronizationStamp(input.getFile());
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think I don't understand the idea behind this change.
Line 425 takes the modificationStamp and stores it as a local synchronizationStamp. Now we try to read getSynchronizationStamp(someIFile) which does

if (element instanceof IFileEditorInput) {
	FileInfo info= (FileInfo) getElementInfo(element);
	if (info != null)
		return info.fModificationStamp;
}
return super.getSynchronizationStamp(element);

it'll always go into the super.getSynchronizationStamp() which returns 0.
Something doesn't really work for me with this code.
Did you mean to call computeModificationStamp instead of getSynchronizationStamp.

If so: There must be something wrong with the tests. They should have caught this.
If not: Please explain (possibly again) in full detail what the idea is.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You're right.. It could never be synchron, the file was reloaded usually. I guess I've to to store the time when the file was loaded and reloaded and compare with isSynchron.. Otherwise it will just compare the real files modification time stamp and the IResource modification timestamp.. But in this case the editor content has to be synchron, not just the file.

I will add a few more asserts to check for isSynchronized() true/false, then we should be more secure with the tests.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I've checked why this override was introduced, but then it doesn't makes a sense to keep this override.
It was introduced to fix the the saving of files which are out of sync: https://bugs.eclipse.org/bugs/show_bug.cgi?id=343894. So If i remove this override, we will have the same problem back, but if I keep my changes in the XtextEditor class and call performSave(true, progressMonitor) .. (example from TextEditor which also forces the save)

So might be someone has fixed the first bug in a wrong place and caused the unsync content problem.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Removed the isSynchronized() overriden method, adjusted the performSave to force override .. Added more assertions to test.

@mehmet-karaman mehmet-karaman force-pushed the fix_xtext_editors_sync_ext_editing branch from be83d08 to 64275cb Compare September 2, 2024 15:12
@@ -327,6 +329,12 @@ public void doSave(IProgressMonitor progressMonitor) {
callback.afterSave(this);
}

@Override
protected void performSave(boolean overwrite, IProgressMonitor progressMonitor) {
super.performSave(true, progressMonitor);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is not something we can merge. Always enforcing an overwrite seems to be the wrong attack vector for the problem at hand.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In comparison to the functionality of the org.eclipse.ui.editors.text.TextFileDocumentProvider we see that they force overwrite in the method org.eclipse.ui.editors.text.TextFileDocumentProvider.createFileFromDocument(IProgressMonitor, IFile, IDocument) .

Also JDT forces the write if the file is synchronized..
org.eclipse.jdt.internal.ui.javaeditor.CompilationUnitDocumentProvider.commitWorkingCopy(IProgressMonitor, Object, CompilationUnitInfo, boolean)

I've adjusted my solution similar to JDT .. I hope this is fine now.

@mehmet-karaman mehmet-karaman force-pushed the fix_xtext_editors_sync_ext_editing branch 2 times, most recently from 41b8718 to 9ef866f Compare September 3, 2024 12:01
@memin20
Copy link

memin20 commented Sep 3, 2024

I am now on my private MacBook and try to understand why it fails constantly on macOS but works on windows and ubuntu..

This bugfix contains the fix for unsync workspace files wrt. the file
system (see the
DirtyStateEditorSupportIntegrationTest.testModifyFileInExternEditor) and
a one liner for beeing able to save the file when the changes were
ignored. (see the ignored test case
DirtyStateEditorSupportIntegrationTest.testModifyDirtyFileInExternEditor).

The issue for this fix can be found here:
eclipse-xtext#2385
@mehmet-karaman mehmet-karaman force-pushed the fix_xtext_editors_sync_ext_editing branch from 9ef866f to 5fa196f Compare September 3, 2024 21:45
@mehmet-karaman
Copy link
Contributor Author

Added waitForBuild to solve this async problem on Mac.. Seems that the build takes longer on mac machines .. Locally it worked already.

/**
* This test needs a manual button click.
*/
@Ignore
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

When I get this dialog:

Screenshot 2024-09-04 at 08 52 51

I'd expect that Replace editor content sets the editor content to the text in the file. But apparently it'll override the file change on disk. That's not ok.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

really? I can't test It yet, as far as i remember it discards the changes in the editor.. Its not supposed to write anything on the disc..

Copy link

@memin20 memin20 Sep 4, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Btw. You have to klick "ignore file changes" and then the changes in XtextEditor is necessary..

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

really? I can't test It yet, as far as i remember it discards the changes in the editor.. Its not supposed to write anything on the disc..

Yes, I tried this a couple of times and added additional assertions to confirm my observation.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Might be that this is due to the doSave() though. Testing again.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, it is. @iloveeclipse can you confirm that it works for you guys?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It works for me..

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, it is. @iloveeclipse can you confirm that it works for you guys?

Mehmet is in my team, and thanks for asking.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Excellent. So let's move this forward. Thank you for the fix and the patience with the review process @mehmet-karaman

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@szarnekow Thank you for your review and yours hints..

Copy link
Contributor

@szarnekow szarnekow left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Change looks good.

@szarnekow szarnekow merged commit 45d8078 into eclipse-xtext:main Sep 5, 2024
9 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants