Re: [fetchmail-devel] Password prompt on stderr

[Matthias A. <mat...@gm...>]

Am 09.02.2009, 12:58 Uhr, schrieb Prasad J Pandit <pj....@ya...>:

>    Hello folks,
>
> I was using fetchmail-6.3.9 as
>
> $ fetchmail -cu <name> -D na...@my... --ssl mail-server 2> /dev/null
>
> and noticed that the prompt which says: "Enter password for %s@%s" is  
> displayed on stderr instead of steout.

Hi Prasad,

thanks for your effort to contribute.

While undocumented, chances are that this was intentional for some reason.  
I've checked the relevant code, and fetchmail opens /dev/tty from file  
descriptor #2 with fdopen(), which is also stderr, to save/change/restore  
settings such as not echoing the password back to the user.

I wonder if it's more common that people discard stdout (progress) instead  
of stderr (problems) - and I suspect that was the reason to use stderr...

At least, if we go this way, we need to fix the fdopen as well to use  
stdout.

I wonder if "-s" (--silent) is an option that works for you if you want  
less output, rather than discarding 2>/dev/null - it is currently unclear  
to me which ACTUAL problem you try to solve by redirecting stderr to the  
bit bucket. Would you care to enlighten me a bit?

> The attached patch fixes this bug and a warning message encountered  
> while compiling the source.
>
>
> The other warning messages were:
>
> smbutil.c:207: warning: the address of ‘lmRespData’ will always evaluate  
> as ‘true’
> smbutil.c:208: warning: the address of ‘ntRespData’ will always evaluate  
> as ‘true’
>
> that's because both are statically allocated array variables.

Yes. :-( I currently understand too little of NTLM authentication to  
actually touch this code myself. I'd be more comfortable with using an  
external NTLM library. I still need to look at the GSoC 2008 patches to  
add OpenChange support (my fault - haven't had the time, and I feel sorry  
for not yet merging Yangyan's code) - perhaps this is a way to do away  
with our old NTLM code.

-- 
Matthias Andree