Ted Zlatanov <[email protected]> writes: > On Thu, 26 May 2011 09:09:40 -0500 Ted Zlatanov <[email protected]> wrote: > > TZ> OK. Is the Git CLI interaction in `magit-password' the only place I > TZ> need to fix? I don't think Magit does anything else with passwords, > TZ> right? See below about askpass... > > Assuming so, here's an untested patch as an initial idea. It > distinguishes between SSH key passphrases and actual passwords. > > For older versions of auth-source.el we'll use the > `auth-source-user-or-password' function. But please check the general > idea of this patch and see if it makes sense. Also check the HTTP > auth--without the origin URL I don't know what I'm looking up... > > Thanks > Ted > > > 68a69,70 >> (require 'auth-source nil t) >> > 1365,1370c1367,1391 > < (when (or (string-match "^Enter passphrase for key '\\\(.*\\\)': $" > string) > < (string-match "^\\\(.*\\\)'s password:" string)) > < (process-send-string proc > < (concat (read-passwd > < (format "Password for '%s': " > (match-string 1 string)) > < nil) "\n"))))
I've recently pushed to master a modification to this part of the code. Could you integrate with it? Note also that when one is reading "foo's password:" it is actually ssh asking for a password when there is no ssh-key. When it's git that do the asking, it just ask for "Password:" (hence the bug for the original poster of this thread). [...] -- Rémi Vanicat
