git push stopped working

[colt]

Hello, I had to delete the working directory of a project. It should have not been a problem, since I expected to clone it from my repository on gitlab and then start working on it again.

And I did it, but when I tried to push a new commit back to gitlab, I received:

[Ant P.]

That remote URL looks totally wrong, usernames should not go near the end. The semicolons in the output look highly suspicious too. Do this to fix it:

[1clue]

FWIW I've never had luck committing when the project was checked out with http. Are you sure you didn't check it out the first time using ssh?

Try checking out using ssh, and use an rsa key as authentication.

[colt]

[John R. Graham]

Hmm. There's no publicly viewable project on Gitlab (gitlab.com) named asmfrt, so that's a potential problem. Do you have a private commercial arrangement with them?

- John
_________________
I can confirm that I have received between 0 and 499 National Security Letters.

[ct85711]

@John, doing a quick search, I did find the asmfrt repo on gitlab; the full name is asmfrt2 (author being asmf). As far as the address in the most recent posts, is correct as what gitlab says (for both https and ssh), and cloning works for me using the https address (did not try pushing anything).

https://gitlab.com/asmf2/asmfrt

[John R. Graham]

Okay, I figured out what I was doing wrong. Successfully cloned, too, via ssh. Thanks very much for pointing out my mistake.

@colt, check your git configuration. Here's what I have:

[colt]

Here is:

[Hu]

Each locally stored repository is independently configured. You could have two local clones, one configured correctly and one configured incorrectly. Is your latest git config --list output from the non-working clone? What error message is produced when you try to push from that clone? (I see that upthread you provided an error message, but I want to be sure that subsequent changes have not invalidated that message.)

Yes, local commits are kept until you delete all references to them (and even then, they are only removed during repository cleanup). Yes, if you fix your configuration to allow publishing, then you can publish commits made before you fixed your configuration.

[colt]

From the "master", the one that I referred initially and that I wish to push to gitlab:

[ct85711]

[Hu]

Normally, I chastise people who repeatedly use incorrect markup, requiring me to correct post after post. In this case, it's actually convenient that you got it wrong. While fixing your markup, I noticed you have an errant character in your configuration. (Though, to be fair, I thought to look for it because of the phrasing of the error message, so I might have caught it even if you had used the correct markup.) It is invisible in the forum, but smarter text editors show it.

Good configuration:

[colt]

Thanks Hu, I wasn't using any text editor, I just copied the output of

[Hu]

According to Wikipedia, that is a Byte Order Mark (BOM). All editors should show it, at least when it is not at the start of the file, but arguably even then. I don't know why both of your preferred editors failed to show it. It got there because you put it there, likely by pasting from some source which had it. Since it seems to be invisible in browsers too, you probably didn't realize it was there when you copied it from a webpage. Since your editors also fail to show it, you didn't see it after you pasted it, either. Looking back, it looks like you probably got it by copying from Ant P.'s post. Indeed, checking that post in a text editor shows that the bad character is hiding in there. So then the question becomes, where did he get that text and how did it get into his post?

[Ant P.]

Ehh... I stuck it in manually to stop the forums automatically turning that text into an invalid link. Didn't think it'd come through copy+paste, but that just reinforces the point that you shouldn't blindly copy and paste random commands off the internet.

Whoops.

[Hu]

Indeed. When I need to inhibit matching, I use a zero-length formatting directive (bold, italic, underline - whichever is not busy in the text at the time). This issue reminds me of LWN: A look at terminal emulators, part 1, which discussed in part some of the amazingly bad things browsers do to make pasting into terminals even more dangerous than people might expect. Among other things, there are sneaky ways to include arbitrary text in the selection that does not render on screen, such as: