bug#35622: Ran into a bug in the new graphical installer on WiFi setup

bug#35622: Ran into a bug in the new graphical installer on WiFi setup

[Hugo Saavedra]

[-- Attachment #1: Type: text/plain, Size: 639 bytes --]

Hello Guix team,

Thanks for your work on GuixSD. I was excited to try out the new graphical
installer, but ran into a bug while setting up WiFi.

I've uploaded photos of the stacktrace for you to take a look at. This is
running on a Dell Inspiron 11 3000 series.

https://imgur.com/a/qcwgNXr

Also, upon clicking "OK" I'm taken back to the original menu of the
installer. Going through the choices again, the installer cannot detect a
hard drive as it could before.

Let me know if I can help in any way or if you need more information.


-- 
*Hugo Saavedra*
Listen Systems <http://listen.systems>
c: 818-356-6664
<http://listen.systems>

[-- Attachment #2: Type: text/html, Size: 1978 bytes --]

bug#35622: Ran into a bug in the new graphical installer on WiFi setup

[Mark H Weaver]

Hi Hugo,

Hugo Saavedra <hm@listen.systems> writes:

> Thanks for your work on GuixSD. I was excited to try out the new
> graphical installer, but ran into a bug while setting up WiFi.

I'm sorry to hear it.  Thanks very much for the report.

> I've uploaded photos of the stacktrace for you to take a look at. This
> is running on a Dell Inspiron 11 3000 series.
>
> https://imgur.com/a/qcwgNXr

From the backtrace, I see that 'string-null?' was applied to #f, and I
guess it was the 'string-null?' called from the 'wifi-services'
procedure in (gnu installer newt wifi), here:

  (define (wifi-services)
    "Return all the connman services of wifi type."
    (let ((services (connman-services)))
      (filter (lambda (service)
                (and (string=? (service-type service) "wifi")
                     (not (string-null? (service-name service)))))
              services)))

It seems that one of the services returned by (connman-services) had #f
as its 'service-name'.  The backtrace includes a (truncated) display of
the service in question:

  #<<service> name: #f type: "wifi" path: "wifi_4cbb58…>

Looking at 'connman-services', it appears that in this case, the 'keys',
as returned by 'parse-keys' in (gnu installer connman), did not have a
"Name" association, or else its right-hand side was #f.

It would be good if someone more familiar with this code would
investigate further.

      Thanks,
        Mark

bug#35622: Ran into a bug in the new graphical installer on WiFi setup

[Ludovic Courtès]

Hello,

Mark H Weaver <mhw@netris.org> skribis:

> Hugo Saavedra <hm@listen.systems> writes:
>
>> Thanks for your work on GuixSD. I was excited to try out the new
>> graphical installer, but ran into a bug while setting up WiFi.
>
> I'm sorry to hear it.  Thanks very much for the report.
>
>> I've uploaded photos of the stacktrace for you to take a look at. This
>> is running on a Dell Inspiron 11 3000 series.
>>
>> https://imgur.com/a/qcwgNXr
>
> From the backtrace, I see that 'string-null?' was applied to #f, and I
> guess it was the 'string-null?' called from the 'wifi-services'
> procedure in (gnu installer newt wifi), here:

Could you post the image here?  imgur.com says it’s “over capacity” and
I can’t see the image right now.

Ludo’.

bug#35622: Ran into a bug in the new graphical installer on WiFi setup

[Hugo Saavedra]

[-- Attachment #1: Type: text/plain, Size: 1142 bytes --]

Sorry about that -- I just posted them on my own server here
https://topologi.es/files/d/5e38edf49f744b099d70/?p=/&mode=grid

On Wed, May 8, 2019 at 7:07 AM Ludovic Courtès <ludo@gnu.org> wrote:

> Hello,
>
> Mark H Weaver <mhw@netris.org> skribis:
>
> > Hugo Saavedra <hm@listen.systems> writes:
> >
> >> Thanks for your work on GuixSD. I was excited to try out the new
> >> graphical installer, but ran into a bug while setting up WiFi.
> >
> > I'm sorry to hear it.  Thanks very much for the report.
> >
> >> I've uploaded photos of the stacktrace for you to take a look at. This
> >> is running on a Dell Inspiron 11 3000 series.
> >>
> >> https://imgur.com/a/qcwgNXr
> >
> > From the backtrace, I see that 'string-null?' was applied to #f, and I
> > guess it was the 'string-null?' called from the 'wifi-services'
> > procedure in (gnu installer newt wifi), here:
>
> Could you post the image here?  imgur.com says it’s “over capacity” and
> I can’t see the image right now.
>
> Ludo’.
>


-- 
*Hugo Saavedra*
Listen Systems <http://listen.systems>
c: 818-356-6664
<http://listen.systems>

[-- Attachment #2: Type: text/html, Size: 2939 bytes --]

bug#35622: Ran into a bug in the new graphical installer on WiFi setup

[Ludovic Courtès]

[-- Attachment #1: Type: text/plain, Size: 1956 bytes --]

Hello,

Mark H Weaver <mhw@netris.org> skribis:

>From the backtrace, I see that 'string-null?' was applied to #f, and I
> guess it was the 'string-null?' called from the 'wifi-services'
> procedure in (gnu installer newt wifi), here:
>
>   (define (wifi-services)
>     "Return all the connman services of wifi type."
>     (let ((services (connman-services)))
>       (filter (lambda (service)
>                 (and (string=? (service-type service) "wifi")
>                      (not (string-null? (service-name service)))))
>               services)))
>
> It seems that one of the services returned by (connman-services) had #f
> as its 'service-name'.  The backtrace includes a (truncated) display of
> the service in question:
>
>   #<<service> name: #f type: "wifi" path: "wifi_4cbb58…>
>
> Looking at 'connman-services', it appears that in this case, the 'keys',
> as returned by 'parse-keys' in (gnu installer connman), did not have a
> "Name" association, or else its right-hand side was #f.

I’ve tried “connmanctl services xyz” on the bare metal with an actual
WiFi device.  For me there’s always a “Name = something” property, and
the “something” appears to be the SSID of the access point.

Could it be that the access point does not advertise an SSID, and thus
its “Name” property is the empty string or is missing altogether?

It could be that changing the ‘parse-keys’ regexp as shown below would
solve the problem for cases where “connmanctl services xyz” writes
literally:

  Name = 

WDYT, Mathieu?

Hugo, would it be an option for you to (1) boot the installation image,
and (2) to grab the output of this command:

  for s in $(connmanctl services | cut -c 25- | grep wifi) ; do connmanctl service $s ; done

Note that it will provide information about the WiFi networks around
you, which you may or may not want to share publicly.

Thanks,
Ludo’.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Type: text/x-patch, Size: 549 bytes --]

diff --git a/gnu/installer/connman.scm b/gnu/installer/connman.scm
index ef8cca3952..b6e3dfd909 100644
--- a/gnu/installer/connman.scm
+++ b/gnu/installer/connman.scm
@@ -170,7 +170,7 @@ to be translated."
 
 Return the corresponding association list of '((KEY . VALUE) (KEY2 . VALUE2)
 ...)  elements."
-  (let ((key-regex (make-regexp "([^ ]+) = ([^$]+)")))
+  (let ((key-regex (make-regexp "([^ ]+) = ([^$]*)")))
     (map (lambda (key)
            (let ((match-key (regexp-exec key-regex key)))
              (cons (match:substring match-key 1)

bug#35622: Ran into a bug in the new graphical installer on WiFi setup

[Ludovic Courtès]

Ludovic Courtès <ludo@gnu.org> skribis:

> I’ve tried “connmanctl services xyz” on the bare metal with an actual
> WiFi device.  For me there’s always a “Name = something” property, and
> the “something” appears to be the SSID of the access point.
>
> Could it be that the access point does not advertise an SSID, and thus
> its “Name” property is the empty string or is missing altogether?

I tried to reproduce that in a VM by running hostapd and using the neat
‘mac80211_hwsim’ kernel module.

However, hostapd’s config file doesn’t allow empty SSIDs.  So I set the
SSID to “ ” (a single space), but that’s correctly handled by the
installer.

Ludo’.