GIT/SSH_ASKPASS used for username input in https:// URLs

Andreas Krey· May 6, 2019, 13:47 UTC · lore

Hi everyone,

there is an interesting wart around prompt.c - PROMPT_ECHO isn't used in invoking an external helper program.

Thus, if I clone something on https (which requires auth for that), and have SSH_ASKPASS set, I will get two GUI *password* prompts, even though the first one will indicate in the title that it *is* asking for the username.

But basically, given the trivial protocol of SSH_ASKPASS, there doesn't seem to be any way to properly fix this without support from any desktop environment that sets SSH_ASKPASS. :-(

The best I can currently tell my users is to use the
   [credentials "https:/..."]
configuration to fix the username.

But if we can't fix this, maybe we can at least point this out in the docs?

- Andreas

-- 
"Totally trivial. Famous last words."
From: Linus Torvalds <torvalds@*.org>
Date: Fri, 22 Jan 2010 07:29:21 -0800

Jeff King· May 7, 2019, 07:36 UTC · re: Andreas Krey · lore

Re: GIT/SSH_ASKPASS used for username input in https:// URLs

On Mon, May 06, 2019 at 03:47:18PM +0200, Andreas Krey wrote:

Show 11 quoted lines

> there is an interesting wart around prompt.c -
> PROMPT_ECHO isn't used in invoking an external helper program.
> 
> Thus, if I clone something on https (which requires auth for that),
> and have SSH_ASKPASS set, I will get two GUI *password* prompts,
> even though the first one will indicate in the title that it *is*
> asking for the username.
> 
> But basically, given the trivial protocol of SSH_ASKPASS, there
> doesn't seem to be any way to properly fix this without support
> from any desktop environment that sets SSH_ASKPASS. :-(

Yep. If you haven't seen it, there's more discussion in this recent thread:

  https://public-inbox.org/git/20190429234028.GA24069@sigill.intra.peff.net/

> The best I can currently tell my users is to use the
>    [credentials "https:/..."]
> configuration to fix the username.

Yes, that works. Or a credential helper could prompt with a better interface (but AFAIK, the only one written that does so is the Windows one).

> But if we can't fix this, maybe we can at least point this out in the
> docs?

Sounds like a good suggestion. Where would you look for it? In the description of GIT_ASKPASS / core.askpass?

-Peff