Are code suggestions broken in Bitbucket 7.4?

[markgillespien]

Developers have started complaining that the newly introduced suggestions feature doesn't work anymore, and everyone gets the following error when trying to use the suggestion.

"We can't apply this suggestion to the source branch. You'll need to make this change manually."

(which is less than helpful, it's a very wordy "something went wrong")

[Bryan Turner]

@markgillespien

I'd recommend opening a support request for this. Our automated test suite for suggestions is all green, and we're unable to reproduce any issues with suggestions internally. If you've recently upgraded, that may be related. That sort of information is better exchanged on a support ticket, rather than here on Community. Once we resolve your issue we can update this question with any relevant details in case someone else happens to hit whatever issue you're hitting.

Best regards,
Bryan Turner
Atlassian Bitbucket

[mark_gillespie]

Sadly our SEN expired 2 weeks ago, and we aren't renewing,

I did try updating GIT to latest supported, and rebooting the server, but it's not helped.  

[Bryan Turner]

I'm interested in helping, but realistically without logs and additional information about your instance there's not a lot I can do. If suggestions fail to apply in a way that shows that popup, there should be an error/stack trace in the logs on the server with more details about the failure.

Your title indicates you're on 7.4 now, but have you recently upgraded? If so, what version did you start out with? Something 6.x, perhaps? When this issue happens for developers, is it in pull requests that were open before the upgrade, or does it also happen in new pull requests created after? (Given 7.4 was only released recently, it seems like there must have been some upgrade involved.)

Unfortunately, here on this public issue isn't a great place to exchange details like this. If you want to ping me directly via email, I'm happy to see if I can get you sorted.

Best regards,
Bryan Turner
Atlassian Bitbucket

[mark_gillespie]

We track versions quite closely, as there are always nice additions (thanks!!), so we came from 7.3 (which did work to my knowledge).

I did run the system log inspector but it didn't come back with any errors, I will take a look the logs and see if I can see anything.

In my mind, there are two possible things that we did within the timeframe it started that could be causing this:

• Upgrade to 7.4
• Change Default Merge Strategy to Squash on Merge.

I don't think it's the later, as we have some other projects that have always been squash on merge, and they didn't have any trouble.  The problem for us appears across all repos and projects.

Looking at the logs, I am seeing alot of these:

2020-08-11 09:37:51,235 INFO [drift:thread-4] c.a.s.i.p.c.d.DriftCommentUpdateProcessor 572:2495@11: Comment 78696 is already in the current diff (617864c71166d55793f43439443f5bebfb2f79bb, d2f70358852ac9a3574cf5cf0590d4a787375feb) and will not be drifted
2020-08-11 09:37:51,235 INFO [drift:thread-4] c.a.s.i.p.c.d.DriftCommentUpdateProcessor 572:2495@11: Comment 78698 is already in the current diff (617864c71166d55793f43439443f5bebfb2f79bb, d2f70358852ac9a3574cf5cf0590d4a787375feb) and will not be drifted
2020-08-11 09:37:51,235 INFO [drift:thread-4] c.a.s.i.p.c.d.DriftCommentUpdateProcessor 572:2495@11: Comment 78707 is already in the current diff (617864c71166d55793f43439443f5bebfb2f79bb, d2f70358852ac9a3574cf5cf0590d4a787375feb) and will not be drifted
2020-08-11 09:37:51,235 INFO [drift:thread-4] c.a.s.i.p.c.d.DriftCommentUpdateProcessor 572:2495@11: Comment 78703 is already in the current diff (617864c71166d55793f43439443f5bebfb2f79bb, d2f70358852ac9a3574cf5cf0590d4a787375feb) and will not be drifted
2020-08-11 09:37:51,235 INFO [drift:thread-4] c.a.s.i.p.c.d.DriftCommentUpdateProcessor 572:2495@11: Comment 78702 is already in the current diff (617864c71166d55793f43439443f5bebfb2f79bb, d2f70358852ac9a3574cf5cf0590d4a787375feb) and will not be drifted
2020-08-11 09:37:51,235 INFO [drift:thread-4] c.a.s.i.p.c.d.DriftCommentUpdateProcessor 572:2495@11: Comments have already been drifted (617864c71166d55793f43439443f5bebfb2f79bb, d2f70358852ac9a3574cf5cf0590d4a787375feb) when processing 1 drift request(s):

[mark_gillespie]

I figured out the problem.  It's related to 3rd party plugin YACC , and the latest version (1.25).  Reverting to the previous version (1.24) and it's working as expected.

The changes in that version were related to checking service accounts.  Unticking the setting in YACC to exclude service accounts still didn't fix this, only downgrading did.

Thanks for your offer of help in getting to the bottom of things.

Raised a bug with the vendor.

[Bryan Turner]

I'm glad you were able to get to the bottom of it.

Just as an FYI, the drift messages are nothing to be concerned about (which is why they're logged at INFO level instead of something more serious); they just indicate that the system went to do some background housekeeping for a given pull request and found that it was already done.

Best regards,
Bryan Turner
Atlassian Bitbucket

[Alexey Remnev]

Hi Bryan.

I'm one of maintainer of mentioned plugin (YACC), which uses repository / global hooks to control the contents of commit message and reject changes if message is not valid.

In version 1.25 we have added checking of changes made from Bitbucket UI and I found that:

1. Bitbucket 7.4.0 Apply suggestion feature produces FILE_EDIT hook event, which fires hook and displays error when hook vetoes this change, but it doesn't display reject details, which doesn't allow developer to understand the reasons of reject:
   
   
   
   If you edit file from source viewer, details message is displayed properly:
   
    
   
   This doesn't seem convenient, so it would be great to see similar message when you try to apply suggestion from pull request comments.
   
   
2. Moreover, if applying suggestion is rejected once, it is cached then and it's not possible to repeat an attempt applying it. Again, committing change from source editor allows to edit and commit change if it was rejected before - would be nice to have a consistence here too.

Thanks.

[Alexey Remnev]

Deleted duplicate.

[Maciej Adamczak]

Thanks for such a good investigation @Alexey Remnev ! I will create a bug ticket in our tracker and forward all of your findings there.

https://jira.atlassian.com/browse/BSERV-12528

Maciej

Bitbucket Developer

[Steven Rhodes]

@Alexey Remnev Where can we get the v1.25 (or v1.26) YACC? It seems they don't exist in the Marketplace anymore.

[Tara English-Sweeney]

@Maciej Adamczak - My team is struggling with this exact issue. I see it is in your long term backog since August and updated in October. Do you have any updates as to when this might be fixed? Thanks!

[Maciej Adamczak]

@Tara English-Sweeney I would recommend contacting the support team and also commenting on the mentioned Jira issue: https://jira.atlassian.com/browse/BSERV-12528

[Alexey Remnev]

Hi all, 

Looks like this issue was fixed in release 7.14 of Bitbucket DC.

[markgillespien]

Whilst the vendor was keen to remove the free version from the marketplace, and leave users high and dry,  I still have a copy of the JAR.  PM me if anyone needs it.   Hopefully the vendor will understand the need to keep the last free version, and not brute force users down the paid route, as It's too easy to click the button in Bitbucket and think you are finding out about costs, when infact it installs the new version that only works for 30 days before needing buying.

I have no issue charging, my issue is the forcing of it.