winmerge | Open Source differencing and merging tool | CSV Processing library

There is no merge.commit option: see the source code, where you can see which options are handled, and observe the lack of a merge.commit option. There was no such option in 2014 either (clone and inspect the Git source to verify this, if you like).

The merge.ff option has been around since then, but merge.commit has, as far as I can tell, never existed. You can of course run a command other than git merge to start merging, and by doing that, you can make this command supply --no-commit as an explicit argument. Note, however, that you cannot alias Git built-ins, so there is no way to make a Git alias that inserts --no-commit when you run git merge. Run git mymerge and make that a Git alias, or run mymerge and make that a shell alias, but do not run git merge.

(Your other option is of course to set branch.name.mergeoptions for every branch name.)

Deleted the project (only from eclipse, don't check to option to delete from disk). Then reimport in workspace solved the problem.

This solution is provided by @Squashman

The problem was that in the same folder I had another batch file names powercfg.bat, so this loop

I tested with default msbuildArgs in my .Net Core Web Application, and everything works fine.

In the meantime I've found a workaround: I remove the references from the project and add them again, and it seems to work. Apparently there's some instability in Visual Studio reference handling.

It sounds like you have a package installed per user (Installation Context and properties: MSIINSTALLPERUSER, ALLUSERS). Then it probably caches in that per user location (user profile) you specified.

Here is a quick scan of your package estate for per user installations. Just save to desktop and run (for example PackageScan.vbs). There will be a message box for each per user installation (if any):

To exclude both extensions in the same filter you can use:

As a GUI application WinMerge probably doesn't have any signaling to let PowerShell know the merge is completed. It's not an API. Any of the wait approaches described so far will probably fail because it's waiting for the process to close which, by your design will never happen.

2 things to think about:

1. If WinMerge has a log file you could monitor the log file for some kind of job completion message. You can write the wait routine looking for that line/text.
2. If there is a log functionality, perhaps use the log to review the jobs after the fact. This may negate the need to leave the interface open, which would solve the wait problem.

Update Per Comments & Edits:

I was wrong I have used WinMerge before as part of Tortoise SVN. Though I prefer Beyond Compare.

I don't think you need to verify the files after RoboCopy. It's reliable enough that corruption is truly an outside risk. RoboCopy uses standard Windows APIs to copy files, and it's doubtful the same skeptisism would accompany a GUI based copy. Understandably some skeptisism may result from the confusing array of options. However, that complexity revolves around what to copy, logging etc..., once RoboCopy decides to copy a file it should do so reliably or error. The copy itself should be considered separate and apart from any uneasyness you may have over the tool's complexity. And, crafting an appropriate Robocopy command is usually a front end effort inclusive of testing, and not often followed by such verifications.

Note: You can also use RoboCopy a second time, to either refresh the data or using the /L option to see if anything was missed or since changed.

Note: When a file copy is in progress RoboCopy sets any new destination file's modified time to 01/01/1980 and only changes it to match the source when the copy is complete. If a file is different by size or modified time, they too aren't changed until the transfer is complete. This is how RoboCopy can be restarted and still find / overwrite incomplete copies.

There is at least one hypothetical situation that could cause corruption. When a device, for example a WAN accelerator gets in between the source and destination. Hypothetically, RoboCopy could think everything is fine based on errant or corrupt responses from that device.

So, dealing with then unlikly risk of file corruption. Your original idea of using Get-FileHash sounds appropriate. However, note that the hash is only performed on file contents. Files with equivalent content but divergent meta-data like name & date will return the same hash. That's quite different than how RoboCopy senses and reacts to file differences which very much relies on name, date & size.

Note: Depending on options RoboCopy may consider other meta-data; attributes & ACLs come to mind.

Without Writing a huge script to calculate and compare hashes across a large set of files you can simply run something like below on the source & destination:

As far as I can see - there is no merge.commit option that you can set, only merge.ff. See https://git-scm.com/docs/git-merge#_configuration.

What you can do is declare an alias that runs merge with the wanted option