[Conkeror] Bug: minibuffer history behaves strangely

[marting at gmx.ch]

I've just reported the following bug on conkeror.org:

(A): 
1. In conkeror, type in 'M-x conkeror-version' ((almost) any
 other interactive command will do as well) and hit ENTER.

2. Type in 'M-x' followed by some characters, say 'qwert'.

3. Type in 'M-p' (minibuffer-history-previous) to access the previous
command, i.e. 'conkeror-version'.

4. Type in 'M-n' (minibuffer-history-next) and notice that the text in
the minibuffer still reads 'conkeror-version', i.e. the input 'qwert'
is irretrievably lost.


(B): 
1. In conkeror, type in 'M-x conkeror-version' ((almost) any
other interactive command will do as well) and hit ENTER.

2. Type in 'M-x'.

3. Type in 'M-n' (minibuffer-history-next) and notice that the text in
the minibuffer now reads 'conkeror-version'.

Notice that both (A) and (B) can be reproduced in a different
minibuffer state too, e.g. by typing 'g' instead of 'M-x'.

I'm not sure everyone agrees but I think conkeror should behave
differently in the two situations above. Namely, after (A).4. I would
expect the minibuffer to contain the text 'qwert', while after
(B).3. it should still be empty (since 'conkeror-version' is the
previous, not the next history entry). That said, I can very well live
with (B) but find (A) quite irritating since (in Emacs) I'm accustomed
to look something up in the minibuffer history while typing in a
command.

Additionally I sometimes experience empty entries in the minibuffer
history; this has happened at least with the 'reload' as well as with
the 'backward-line' command. However, I am at the moment not able to
reproduce this behaviour. Has anyone else ever experienced it?

--
Martin