Sending bug reports without setting up email

Sending bug reports without setting up email

[Lars Ingebrigtsen]

We discussed this before, and talked about web forms etc, but it
occurred to me tonight (while trying to fall asleep) how easy it would
be just to use smtpmail.el:

If Emacs detects that email hasn't been configured when trying to send a
bug report, Emacs could just contact the MX for debbugs.gnu.org directly
and submit the report.

Now, that MX ((which happens to be "debbugs.gnu.org") only listens at
port 25, and some ISPs block outgoing traffic to port 25, so it wouldn't
always work.  Is there any chance that we could have that Exim instance
listen on port 587, too?  (The mail submission port.)

In that case, `C-c C-c' when sending the bug report would

1) If mail is set up, just send normally
2) If not, see whether it can talk to 25/587 on the MX
2b) If there is no good "from" address, ask the user what their email
address is
3) If it can't talk to the MX, continue as now with setting up the mail
transport

I think this would be a nice improvement in the usability of the bug
reporting system...

To look up the MX in a reliable, and non-hardcoded manner, it would be
nice if we could use dns.el.  Would anybody mind if I create a C
interface function to return nsaddr_list (from res_init) to Lisp?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Sending bug reports without setting up email

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> To look up the MX in a reliable, and non-hardcoded manner, it would be
> nice if we could use dns.el.  Would anybody mind if I create a C
> interface function to return nsaddr_list (from res_init) to Lisp?

Paul, if I understand how this works correctly, this means that I would
have to undo parts of this commit:

93bf7d52841c60ffc10e0c9c789a7987812ce55e

That is, put the check for res_init back again.   Would that be OK?

And on Windows (or just Cygwin?) some other code might be necessary...

http://stackoverflow.com/questions/10523828/how-does-gcc-cygwin-get-the-dns-server/10773280#10773280

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Sending bug reports without setting up email

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Date: Mon, 09 May 2016 20:19:38 +0200
> Cc: eggert@cs.ucla.edu
> 
> Paul, if I understand how this works correctly, this means that I would
> have to undo parts of this commit:
> 
> 93bf7d52841c60ffc10e0c9c789a7987812ce55e
> 
> That is, put the check for res_init back again.   Would that be OK?
> 
> And on Windows (or just Cygwin?) some other code might be necessary...
> 
> http://stackoverflow.com/questions/10523828/how-does-gcc-cygwin-get-the-dns-server/10773280#10773280

Paul's changes don't affect the native Windows builds, AFAIU, and the
stackoverflow article is explicitly about Cygwin.

Re: Sending bug reports without setting up email

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> Paul's changes don't affect the native Windows builds, AFAIU, and the
> stackoverflow article is explicitly about Cygwin.

Right.  And plain Windows uses a totally different method to get the DNS
addresses?  I googled a bit and found something about
GetAdaptersAddresses...

https://msdn.microsoft.com/en-us/library/windows/desktop/aa365915%28v=vs.85%29.aspx

Which I guess means that this is done in a completely different fashion
in Windows, so I'll let somebody else worry about it.  :-)

Anyway, I've now got a proof of concept thing working on GNU/Linux, but
it'll probably break all other builds without putting the res_init
configure.ac check back in.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Sending bug reports without setting up email

[Paul Eggert]

On 05/09/2016 11:19 AM, Lars Ingebrigtsen wrote:
> That is, put the check for res_init back again.   Would that be OK?

It should be OK to put the res_init check back into configure.ac. 
However, res_init is not thread-safe, and other threads might be doing 
name resolution. Shouldn't Emacs use res_ninit instead?

Re: Sending bug reports without setting up email

[daniel sutton]

[-- Attachment #1: Type: text/plain, Size: 767 bytes --]

Another easy solution to this would be to place a copy of the email body on
the clipboard and allow the user to paste in whatever email client they
prefer. As long as you include the destination email in the email body,
this would be easy to get and display a message that the email is on the
system clipboard, you could get this nice feature very cheaply.

On Tue, May 10, 2016 at 10:02 AM, Paul Eggert <eggert@cs.ucla.edu> wrote:

> On 05/09/2016 11:19 AM, Lars Ingebrigtsen wrote:
>
>> That is, put the check for res_init back again.   Would that be OK?
>>
>
> It should be OK to put the res_init check back into configure.ac.
> However, res_init is not thread-safe, and other threads might be doing name
> resolution. Shouldn't Emacs use res_ninit instead?
>
>
>

[-- Attachment #2: Type: text/html, Size: 1283 bytes --]

Re: Sending bug reports without setting up email

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 531 bytes --]

On Wed, May 11, 2016 at 4:55 PM daniel sutton <danielsutton01@gmail.com>
wrote:

> Another easy solution to this would be to place a copy of the email body
> on the clipboard and allow the user to paste in whatever email client they
> prefer.
>

"C-x h M-w" already does the same thing but better. It copies the subject,
email, body, everything. That way you do not need to copy 3 different
things when emailing from a different email client. I use the C-x h M-w
approach when emailing bug reports via Gmail.
-- 

-- 
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 905 bytes --]

Re: Sending bug reports without setting up email

[Eli Zaretskii]

> Date: Wed, 11 May 2016 15:55:42 -0500
> From: daniel sutton <danielsutton01@gmail.com>
> Cc: Lars Ingebrigtsen <larsi@gnus.org>, emacs-devel <emacs-devel@gnu.org>
> 
> Another easy solution to this would be to place a copy of the email body on the clipboard and allow the user to
> paste in whatever email client they prefer. As long as you include the destination email in the email body, this
> would be easy to get and display a message that the email is on the system clipboard, you could get this nice
> feature very cheaply.

We already have this supported.

Re: Sending bug reports without setting up email

[Eli Zaretskii]

> From: Kaushal Modi <kaushal.modi@gmail.com>
> Date: Wed, 11 May 2016 21:00:54 +0000
> Cc: Lars Ingebrigtsen <larsi@gnus.org>, emacs-devel <emacs-devel@gnu.org>
> 
>  Another easy solution to this would be to place a copy of the email body on the clipboard and allow the
>  user to paste in whatever email client they prefer. 
> 
> "C-x h M-w" already does the same thing but better.

There's no need to do anything like that, report-emacs-bug already
offers an alternative to put the entire message into the clipboard.

Re: Sending bug reports without setting up email

[Kaushal Modi]

[-- Attachment #1: Type: text/plain, Size: 498 bytes --]

On Wed, May 11, 2016 at 5:06 PM Eli Zaretskii <eliz@gnu.org> wrote:

> There's no need to do anything like that, report-emacs-bug already
> offers an alternative to put the entire message into the clipboard.
>

I believe you are referring to C-c M-i while in the *message* buffer? That
approach does not work for me as I copy the *message* buffer text across
VNC from linux to Windows to send email via Gmail. xdg-open/gvfs-open does
not work on the system where I run emacs.
-- 

-- 
Kaushal Modi

[-- Attachment #2: Type: text/html, Size: 807 bytes --]

Re: Sending bug reports without setting up email

[Eli Zaretskii]

> From: Kaushal Modi <kaushal.modi@gmail.com>
> Date: Wed, 11 May 2016 21:17:36 +0000
> Cc: larsi@gnus.org, eggert@cs.ucla.edu, danielsutton01@gmail.com, 
> 	emacs-devel@gnu.org
> 
>  There's no need to do anything like that, report-emacs-bug already
>  offers an alternative to put the entire message into the clipboard.
> 
> I believe you are referring to C-c M-i while in the *message* buffer?

No, I mean type "mail client" at the prompt shown by Emacs when you
type "C-c C-c" to send the report.

(We are still talking about report-emacs-bug, aren't we?)

Re: Sending bug reports without setting up email

[Lars Ingebrigtsen]

Paul Eggert <eggert@cs.ucla.edu> writes:

> On 05/09/2016 11:19 AM, Lars Ingebrigtsen wrote:
>> That is, put the check for res_init back again.   Would that be OK?
>
> It should be OK to put the res_init check back into
> configure.ac. However, res_init is not thread-safe, and other threads
> might be doing name resolution. Shouldn't Emacs use res_ninit instead?

Yes, res_ninit would be better, I think.  But is res_ninit supported
less/more than res_init on non-GNU/Linux systems, I wonder?  I tried
googling for this briefly, but was unable to say...

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Sending bug reports without setting up email

[Paul Eggert]

On 07/20/2016 03:29 PM, Lars Ingebrigtsen wrote:
> Yes, res_ninit would be better, I think.  But is res_ninit supported
> less/more than res_init on non-GNU/Linux systems, I wonder?  I tried
> googling for this briefly, but was unable to say...

If memory serves, res_ninit was introduced by BIND 8.2 in the previous 
millennium and nowadays any system that has the deprecated res_init 
function should also have res_ninit. For what it's worth res_ninit is on 
the oldest platform I have easy access to (Solaris 10, released 2005). 
On systems that lack res_ninit we could punt and pretend there are no 
name servers.

Sorry, I've lost track of why you wanted direct access to the list of 
name servers. Something about looking up MX records for gnu.org when 
half-a-dozen other bits of code don't work?  For something that esoteric 
maybe it'd be simpler to hard-code the IP addresses.

Re: Sending bug reports without setting up email

[Lars Ingebrigtsen]

Paul Eggert <eggert@cs.ucla.edu> writes:

> Sorry, I've lost track of why you wanted direct access to the list of
> name servers. Something about looking up MX records for gnu.org when
> half-a-dozen other bits of code don't work?

Yes, for sending bug reports without setting up mail...

> For something that esoteric maybe it'd be simpler to hard-code the IP
> addresses.

It's also useful slightly more in general.  For instance, the way to
upload blog posts to wordpress.com is to make a HTML email with embedded
images.  Their MTA accepts huge emails, while your local MTA is unlikely
to, so sending it directly to their MTA is the way to go.

And you could also see somebody who is very security conscious, I mean
paranoid, would want to send emails via TLS directly to the recipient's
MTA without going through the NSA, I mean intermediaries.

So there's three use cases.  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Sending bug reports without setting up email

[Paul Eggert]

On 07/20/2016 05:06 PM, Lars Ingebrigtsen wrote:
> So there's three use cases.  :-)

OK. Sorry, I still don't recall the underlying issue, but all three use 
cases are covered by libresolv, no? So perhaps we should be using the 
library rather than reinventing most of its wheel.

Re: Sending bug reports without setting up email

[Lars Ingebrigtsen]

Paul Eggert <eggert@cs.ucla.edu> writes:

> On 07/20/2016 05:06 PM, Lars Ingebrigtsen wrote:
>> So there's three use cases.  :-)
>
> OK. Sorry, I still don't recall the underlying issue, but all three
> use cases are covered by libresolv, no? So perhaps we should be using
> the library rather than reinventing most of its wheel.

Sure, but that's an additional library.  We just need the IP address of
the DNS server (provided by res_{n,}init) to do MX resolution via
dns.el.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Sending bug reports without setting up email

[Paul Eggert]

On 07/20/2016 06:10 PM, Lars Ingebrigtsen wrote:
> Sure, but that's an additional library.  We just need the IP address of
> the DNS server (provided by res_{n,}init)

On some systems res_ninit and res_init are in libresolv, so this 
wouldn't save us anything.

No big deal I suppose. Still, it's better to not reinvent the wheel, as 
a lot of things can go wrong when resolving names.

Re: Sending bug reports without setting up email

[Lars Ingebrigtsen]

Paul Eggert <eggert@cs.ucla.edu> writes:

> On 07/20/2016 06:10 PM, Lars Ingebrigtsen wrote:
>> Sure, but that's an additional library.  We just need the IP address of
>> the DNS server (provided by res_{n,}init)
>
> On some systems res_ninit and res_init are in libresolv, so this
> wouldn't save us anything.
>
> No big deal I suppose. Still, it's better to not reinvent the wheel,
> as a lot of things can go wrong when resolving names.

True.  If libresolv is an option, then that would be a more complete
solution.

I guess it's up to the maintainers to make the decision: A teensy tiny
change in the C layer, but with a possibly less complete solution, or a
larger change, but with a complete solution.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Sending bug reports without setting up email

[Stefan Monnier]

> And you could also see somebody who is very security conscious, I mean
> paranoid, would want to send emails via TLS directly to the recipient's
> MTA without going through the NSA, I mean intermediaries.

Of course, nowadays this is becoming a losing proposition: MTAs only
accept email from the sender's official MTA.  So you (the sender) first
need to send your email (via TLS) to your MTA, and *it* will then send
it (maybe in the clear maybe not) to the recipient's MTA.


        Stefan

Re: Sending bug reports without setting up email

[Lars Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> Of course, nowadays this is becoming a losing proposition: MTAs only
> accept email from the sender's official MTA.  So you (the sender) first
> need to send your email (via TLS) to your MTA, and *it* will then send
> it (maybe in the clear maybe not) to the recipient's MTA.

No, MTAs accept messages for their own recipients, of course.  A few
block incoming connections from non-MTA hosts, but that's pretty rare.

But quite a few ISPs block port 25 (and some block the SMTP submission
port, too) for outgoing traffic.  In that case you have to talk to your
local MTA.

I don't have any numbers, but when travelling the past few years I've
never had any trouble talking to the SMTP submission port on the MTA I
use in Norway, so blocking is pretty rare from hotel wifis and the like,
at least.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no