Make USB modems just work on Guix

When plugging in a USB modem, i.e. a “surf stick” / internet stick

with a SIM card for mobile broadband, setup should be as simple as

WiFi. However, a few things are missing in Guix.

First, the manual claims:

-- Scheme Variable: modem-manager-service-type

[…]

This service is part of ‘%desktop-services’ (*note Desktop

Services::).

But it is not part of %desktop-services. The first attached patch

fixes this oversight.

Now I can type this to get an internet connection:

sudo nmcli connection add type gsm ifname ttyUSB1 apn internet

However, I want to be able to do this setup graphically. I made a

second patch to propagate iso-codes (for mobile broadband country

selection) and mobile-broadband-provider-selection (for selecting

defaults for providers) in nm-connection-editor

(network-manager-applet) and gnome-control-center.

Propagation is ugly though. Should I try patching

network-manager-applet instead? I assume there are other packages

doing such patching that I can imitate.

Now I can configure the modem as Mobile Broadband in a graphical

interface. I cannot do so in nmtui; I believe this modem setup

feature is not included in nmtui.

However, this is not the whole truth. Before I can configure

anything, I have to run

sudo ~/usr/sbin/usb_modeswitch -c ~/Downloads/usb-modeswitch-data-20170806/usb_modeswitch.d/12d1:14fe -v 0x12d1 -p 0x14fe

because my modem by default presents itself as a CD containing Windows

drivers and not as a USB modem. Preferrably, usb-modeswitch and

usb-modeswitch-data would be packaged (as a single package). I can

try that later this week.

Lastly, there are udev rules in usb-modeswitch, I believe, because on

Debian I do not need to run any command; mode switching happens

automatically when plugging in the USB modem. I suppose that means

there should also be a usb-modeswitch-service-type extending udev?

Regards,

Florian

From 5e9b45fe785dee52da2a67fbecbabc4741f85e54 Mon Sep 17 00:00:00 2001

---

gnu/services/desktop.scm | 3 +++

1 file changed, 3 insertions(+)

From d489f805599ea63e2e06727a3201572ee0480a60 Mon Sep 17 00:00:00 2001

for USB modem setup.

mobile-broadband-provider-info.

(gnome-control-center): Ditto.

---

gnu/packages/gnome.scm | 10 ++++++++--

1 file changed, 8 insertions(+), 2 deletions(-)

On Wed, May 08, 2019 at 08:26:07PM +0200, pelzflorian (Florian Pelz) wrote:

I meant mobile-broadband-provider-info.

There is a branch wip-gnome3.30. Should I base the patch on

wip-gnome3.30 or on master?

There are two source URLs, so these two must be two packages, right?

Regards,

Florian

I think back then we thought it would be too big for the installation medium.

How big is it after that?

On Thu, May 09, 2019 at 02:55:04PM +0200, Danny Milosavljevic wrote:

The installation medium uses %desktop-services? I will check, but my

proposal does not concern the installation medium.

Regards,

Florian

OBOn Thu, May 09, 2019 at 03:07:12PM +0200, pelzflorian (Florian Pelz) wrote:

There is no size difference. The patches only affect installed systems.

Regards,

Florian

Hi Florian,

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

The comment is unnecessary, but otherwise LGTM!

I’ve confirmed that there is no size increase:

Would it be an option to avoid propagation by instead substituting

references to these two things in the code of network-manager-applet and

gnome-control-center? WDYT?

Thanks!

Ludo’.

On Thu, May 09, 2019 at 05:24:39PM +0200, Ludovic Courtès wrote:

:) I have attached the patch without the command.

Also attached is my current work-in-progress package for

USB_ModeSwitch. It does *not* work because I cannot figure out how to

use tcl in the package’s inputs field.

#:use-module gives me errors; only (@ (gnu packages tcl) tcl) appears

to work.

I will try finding and imitating an example for such substitutions

tomorrow.

Regards,

Florian

From 827d931eb6b72572f547c87e650780cf50e9e20c Mon Sep 17 00:00:00 2001

---

gnu/services/desktop.scm | 1 +

1 file changed, 1 insertion(+)

On Mon, May 13, 2019 at 10:57:20PM +0200, pelzflorian (Florian Pelz) wrote:

I uploaded

+ (inputs `(("libusb" ,libusb)

+ ("tcl" ,(error tcl:tcl))))

This was a mistake, I mean

+ (inputs `(("libusb" ,libusb)

+ ("tcl" tcl:tcl)))

It still gives me errors when used:

florian@florianmacbook ~/git/guix [env]$ ./pre-inst-env guix install usb-modeswitch:dispatcher usb-modeswitch

Backtrace:

18 (apply-smob/1 #<catch-closure 7aa0a0>)

In ice-9/boot-9.scm:

705:2 17 (call-with-prompt _ _ #<procedure default-prompt-handler (k proc)>)

In ice-9/eval.scm:

619:8 16 (_ #(#(#<directory (guile-user) 893140>)))

In guix/ui.scm:

1747:12 15 (run-guix-command _ . _)

In ice-9/boot-9.scm:

829:9 14 (catch _ _ #<procedure 7fd151e1a910 at guix/ui.scm:703:2 (key c)> _)

829:9 13 (catch _ _ #<procedure 7fd151e1a928 at guix/ui.scm:826:6 (key proc format-string format-args . rest)> _)

In guix/scripts/package.scm:

915:10 12 (_)

In guix/status.scm:

768:4 11 (call-with-status-report _ _)

In guix/scripts/package.scm:

923:14 10 (_)

882:6 9 (process-actions #<store-connection 256.99 cfeea0> _)

In guix/ui.scm:

1096:4 8 (show-manifest-transaction #<store-connection 256.99 cfeea0> _ _ #:dry-run? _)

In srfi/srfi-1.scm:

628:19 7 (mapn ("usb-modeswitch" "usb-modeswitch") (("2.5.2" "2.5.2") ("dispatcher" "out") (#<package usb-modeswitch@2.5.2 gnu/packages/libusb.scm:642 174c8f0> #<package usb-modeswitch@2.5.2 gnu/packages/libusb.scm:642 174c8f0>)) 2)

In guix/ui.scm:

1101:23 6 (_ "usb-modeswitch" "2.5.2" "dispatcher" #<package usb-modeswitch@2.5.2 gnu/packages/libusb.scm:642 174c8f0>)

In guix/packages.scm:

1296:13 5 (package-output #<store-connection 256.99 cfeea0> #<package usb-modeswitch@2.5.2 gnu/packages/libusb.scm:642 174c8f0> _ _)

936:16 4 (cache! #<weak-table 178/223> #<package usb-modeswitch@2.5.2 gnu/packages/libusb.scm:642 174c8f0> ("x86_64-linux" . #t) _)

1254:22 3 (thunk)

936:16 2 (cache! #<weak-table 178/223> #<package usb-modeswitch@2.5.2 gnu/packages/libusb.scm:642 174c8f0> ("x86_64-linux" #f #t) _)

1042:46 1 (thunk)

In gnu/packages/libusb.scm:

655:12 0 (inputs _)

gnu/packages/libusb.scm:655:12: In procedure inputs:

error: tcl:tcl: unbound variable

Regards,

Florian

On Mon, May 13, 2019 at 11:04:43PM +0200, pelzflorian (Florian Pelz) wrote:

This tcl:tcl gets resolved all right elsewhere in the libusb.scm file.

There must be something special about <package> definitions or their

inputs.

Regards,

Florian

Hi Florian,

On Mon, 13 May 2019 23:04:43 +0200

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> wrote:

The problem is there's a use-modules cycle somewhere.

Should be

(inputs `(("libusb" ,libusb)

("tcl" ,tcl:tcl)))

But I guess you meant that anyway.

The problem is there's a use-modules cycle somewhere.

Try putting usb-modeswitch in another module (a new one if necessary).

Also, does it really require tcl? Sounds kinda weird to me.

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

In short, you cannot do that for (gnu packages …) modules.

But if you remove that #:prefix, it’s hopefully fine?

Thanks,

Ludo’.

On Tue, May 14, 2019 at 10:35:55AM +0200, Danny Milosavljevic wrote:

Yes I did mean that with comma, sorry.

Using a new module helps, thank you! Now I also do not need a prefix

tcl: anymore. @Ludo: without that prefix, I got errors because 'zip

was redefined or something, I do not remember now.

Now only packaging issues remain. When I use trivial-build-system,

#:builder

(begin

(use-modules (guix build utils)

(guix packages))

(let ((source (assoc-ref %build-inputs "source"))

(tar (assoc-ref %build-inputs "tar"))

(bzip2 (assoc-ref %build-inputs "bzip2"))

(share-dir (string-append %output "/share")))

(copy-file source "data.tar.bz2")

(invoke (string-append bzip2 "/bin/bzip2") "-d" "data.tar.bz2")

(invoke (string-append tar "/bin/tar") "xvf" "data.tar")

(install-file (string-append "usb-modeswitch-data-"

(package-version this-package)

"/usb_modeswitch.d") share-dir))

#t)))

this package-version call seems not to work (at least not without more

adding more modules). Is this the right approach? Also I forgot, do

I need this #t at the end?

TCL is not needed for command-line mode switching, but there is a TCL

script called usb_modeswitch_dispatcher that gets called by the UDEV

rule and by the systemd service file shipped with USB_ModeSwitch.

This dispatcher automatically modeswitches USB devices that are

plugged in, I believe.

Regards,

Florian

On Mon, May 13, 2019 at 10:57:20PM +0200, pelzflorian (Florian Pelz) wrote:

I have attached the patch again without the comment and with

mentioning modem support in the Desktop services section in the

doc/guix.texi manual. Previously it was only said in the Network

services section that modem-manager-service-type were part of

%desktop-services.

Regards,

Florian

From 061c976e38b8b332602e84ac74717384aab57d56 Mon Sep 17 00:00:00 2001

---

doc/guix.texi | 17 +++++++++--------

gnu/services/desktop.scm | 1 +

2 files changed, 10 insertions(+), 8 deletions(-)

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

Applied. I had applied the desktop.scm yesterday, so I’ve taken just

the texi part here.

Thanks!

Ludo’.

Hi,

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

[...]

That could be a collision with ‘zip’ from (srfi srfi-1) or with ‘zlib’

from (guix licenses), which can both be worked around by doing something

along these lines:

#:use-module ((srfi srfi-1) #:hide (zip))

^

You’re missing a comma (“unquote”) so that ‘package-version’ is called

on the host side; it doesn’t exist on the build side.

#t is unnecessary here.

Just “New file”. Make sure to add the file to gnu/local.mk please.

Please adjust the module name: it should be ‘usb-modeswitch’.

Please make it a full sentence, and ideally expound a bit (info "(guix)

Synopses and Descriptions").

Make sure to remove any leftovers like this comment before sending the

next version. :-)

Please don’t refer to global variables ‘%outputs’ and ‘%build-inputs’.

Instead, take them from the phase’s arguments:

(lambda* (#:key source outputs inputs #:allow-other-keys)

(let* ((out (assoc-ref outputs "out")))

…))

The phase must return #t.

Probably `multi-mode', as expected in Texinfo.

… and @dfn{modes}

Could you send an updated patch?

Is USB_ModeSwitch needed for proper use of ModemManager? I’d say “no”

but I’m unsure.

Thanks for working on it!

Ludo’.

On Wed, May 15, 2019 at 03:01:04PM +0200, Ludovic Courtès wrote:

Thank you very much! Will do later, today or tomorrow.

Only on some modems.

:)

Regards,

Florian

On Thu, May 09, 2019 at 05:24:39PM +0200, Ludovic Courtès wrote:

I have confirmed that the patch attached to this e-mail makes

network-manager-applet’s libnma library load its data from the store

and not the profile. This fixes nm-connection-editor from

network-manager-applet as well as gnome-control-center which

previously required modem-broadband-provider-info and iso-codes to be

installed in the profile.

I am still working on packaging USB_ModeSwitch which some USB modems

require.

Regards,

Florian

From 627e41c4741acd5b91771131a1646bb89b8c270b Mon Sep 17 00:00:00 2001

iso-codes and mobile-broadband-provider-info from the store and not require

them in the profile.

---

gnu/packages/gnome.scm | 22 ++++++++++++++++++++--

1 file changed, 20 insertions(+), 2 deletions(-)

On Thu, May 16, 2019 at 09:50:18AM +0200, pelzflorian (Florian Pelz) wrote:

I hope it would not have been required to increase a revision number.

Regards,

Florian

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

Great. Applied with the changes below; the spacing change is to please

Emacs (C-M-a & co. get confused by opening parentheses at the beginning

of a line.)

Thanks,

Ludo’.

Thank you for your thorough review. Find attached a new version of

the patch. I can now modeswitch my USB modem by doing

./pre-inst-env guix install usb-modeswitch:dispatcher

sudo usb_modeswitch -c ~/.guix-profile/share/usb_modeswitch.d/12d1\:14fe -v 0x12d1 -p 0x14fe

Tring to put the packages in gnu/packages/libusb.scm and not in their

own file gnu/packages/usb-modeswitch.scm again resulted in

[ 11%] LOAD guix/scripts/search.scm

[ 11%] LOAD guix/scripts/gc.scm

[ 11%] LOAD guix/scripts/hash.scm

[ 11%] LOAD guix/scripts/pack.scm

Backtrace:

In ice-9/boot-9.scm:

222:29 19 (map1 (((guix utils)) ((guix store)) ((guix #) # (#)) …))

222:29 18 (map1 (((guix store)) ((guix status) #:select (#)) (#) …))

222:29 17 (map1 (((guix status) #:select (#)) ((guix grafts)) # …))

222:29 16 (map1 (((guix grafts)) ((guix monads)) ((guix #)) (#) …))

222:29 15 (map1 (((guix monads)) ((guix modules)) ((guix #)) (#) …))

222:29 14 (map1 (((guix modules)) ((guix packages)) ((guix #)) # …))

222:29 13 (map1 (((guix packages)) ((guix profiles)) ((guix #)) …))

222:29 12 (map1 (((guix profiles)) ((guix describe)) ((guix #)) …))

222:29 11 (map1 (((guix describe)) ((guix derivations)) ((# #)) …))

222:29 10 (map1 (((guix derivations)) ((guix search-paths)) (#) …))

222:29 9 (map1 (((guix search-paths)) ((guix build-system #)) # …))

222:29 8 (map1 (((guix build-system gnu)) ((guix scripts #)) # …))

222:29 7 (map1 (((guix scripts build)) ((guix self) #:select …) …))

222:29 6 (map1 (((guix self) #:select (make-config.scm)) ((…)) …))

222:29 5 (map1 (((gnu packages)) ((gnu packages bootstrap)) (…) …))

222:29 4 (map1 (((gnu packages bootstrap)) ((gnu packages …) …) …))

222:17 3 (map1 (((gnu packages compression) #:hide (zip)) ((…)) …))

2830:10 2 (resolve-interface (gnu packages compression) #:select _ …)

260:13 1 (for-each #<procedure 52470e0 at ice-9/boot-9.scm:2830…> …)

In unknown file:

0 (scm-error misc-error #f "~A" ("no binding `zip' to h…") …)

ERROR: In procedure scm-error:

no binding `zip' to hide in module (gnu packages compression)

when running make. Then I rebuilt and commented and uncommented and

moved some code and putting the code in gnu/packages/libusb.scm worked

again in ./pre-inst-env. Later it failed again. Danny speculated the

reason was a use-module cycle. I would like confirmation that this is

what a use-module cycle looks like and that I should stop trying to

put usb-modeswitch back in gnu/packages/libusb.scm.

As for the packaging, I hope it is correct to still refer to %outputs

in the trivial-build-system for usb-modeswitch-data.

Should I also make the usb-modeswitch-data package install the license

file of usb-modeswitch-data?

Or should I use gnu-build-system instead of trivial-build-system?

trivial works now though.

The file dispatcher.c of usb-modeswitch is licensed bsd-2 and not

gpl2+. Should I list both licenses or just gpl2+?

I have checked the license of the files of usb-modeswitch-data and

usb-modeswitch by going through each of them with the `less` program

and reading its header while assuming that no license header means the

file has the license described in the README file, by diff’ing the

license files with a license downloaded from gnu.org, and that nothing

looked like a program blob (there only are very small hex strings for

the messages USB_ModeSwitch sends like

MessageContent="55534243123456782400000080000612000024000000000000000000000000"

that seem harmless). Is this the usual procedure?

Currently I also install the UDEV rule shipped with

usb-modeswitch-data; I hope this is correct.

Currently I install the usb_modeswitch_dispatcher that is needed by

the udev rule in a separate output called “dispatcher”. I am unsure

if this usb_modeswitch_dispatcher would need the usb_modeswitch

program from the “out” output. Is it OK when one output only works

when the “out” output is installed too?

What is still missing is adding a service type for usb-modeswitch that

extends the udev service type by the rules file and that calls the

usb-modeswitch dispatcher once every reboot. I can try making one,

but I am not sure if this is the right approach. I am also unsure if

this service should be part of %desktop-services; many modems need it

though and modem-manager-service-type.

Regards,

Florian

I forgot to mention gnu/local.mk in the commit message. Please use

this patch attached here instead.

Regards,

Florian

Hello,

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

You cannot use #:hide on (gnu packages …) modules, for unfortunate and

obscure reasons. The trick would be to hide the other ‘zip’ binding,

the one from (srfi srfi-1).

Yes.

In general we should install license files. If ‘trivial’ works for you,

that’s OK; an additional line to install the license would be welcome,

but not blocking.

I think it’s enough to mark it as ‘gpl2+’ because, IIUC, the combination

of all this is GPLv2+. If you think that needs clarification, you can

just add a comment about the ‘license’ field.

Yes. I often just scan a few source file headers, not necessarily all

of them, but of course doing an exhaustive search is even better!

Yep.

I’d say no. :-) In general, we want “guix install foo:bar” to provide

something that Just Works.

We’ll see that later. :-)

Let me know if you want to send an updated patch if you we should

proceed with the one you sent.

Thanks!

Ludo’.

Thank you for your feedback!

On Fri, May 24, 2019 at 05:37:01PM +0200, Ludovic Courtès wrote:

I did not #:hide anything. The error is reported when (or after?) a

call to “make” loads guix/scripts/pack.scm (which I have not changed)

after putting the packages and required modules into libusb.scm

instead of usb-modeswitch.scm.

I add the license file.

Since nobody is likely to use the USB_ModeSwitch dispatcher.c without

USB_ModeSwitch, I tend to leave it at gpl2+.

I will try substitute*ing the dispatcher script to use /gnu/store

until it just works.

I will send an updated patch tomorrow with a working ”dispatcher“

output.

Regards,

Florian

On Fri, May 24, 2019 at 10:12:10PM +0200, pelzflorian (Florian Pelz) wrote:

Please find attached a patch that no longer propagates

usb-modeswitch-data from usb-modeswitch. I can now modeswitch my USB

modem by running

$ ./pre-inst-env guix install usb-modeswitch:dispatcher

$ lsmod -v #find bus name and device name of modem

[…]

Bus 001 Device 007: ID 12d1:14fe Huawei Technologies Co., Ltd. Modem (Mass Storage Mode)

$ for d in /sys/bus/usb/devices/*/; do echo $d; cat $d/busnum; cat $d/devnum; done

[…]

/sys/bus/usb/devices/1-1/

1

7

[…]

$ sudo usb_modeswitch_dispatcher --switch-mode 1-1 --config-file ~/.guix-profile/etc/usb_modeswitch.conf

I added the --config-file option since by default

usb_modeswitch_dispatcher loads its optional config file from /etc and

a future Guix service type for USB_ModeSwitch should be able to pass a

config file.

The udev rules file in usb-modeswitch-data has lines like:

ATTR{idVendor}=="03f0", ATTR{idProduct}=="002a", RUN+="usb_modeswitch '/%k'"

Maybe I should substitute these by

RUN+="/gnu/store/…-usb-modeswich-2.5.2/usb_modeswitch '/%k'"

but this would introduce a cyclic dependency between usb-modeswitch

and usb-modeswitch-data. Therefore this rules file should rather be

fixed in a future USB_ModeSwitch service type.

Regards,

Florian

On Sat, May 25, 2019 at 01:18:56PM +0200, pelzflorian (Florian Pelz) wrote:

Or should I rather use a patch file instead of substitute*?

Hello,

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

Yay!

Sounds good. Another option would be to install the udev rule as part

of the ‘usb-modeswitch’ package, which makes more sense and would solve

the problem. But anyway, don’t lose your hair over it.

I don’t think it’s needed, is it?

I think it’s OK like this. It does mean Guix’s

‘usb_modeswitch_dispatcher’ has an option that upstream doesn’t provide.

It would be worth proposing it upstream.

Thanks!

Ludo’.

On Sat, May 25, 2019 at 07:44:37PM +0200, Ludovic Courtès wrote:

The udev rules file is shipped with usb-modeswitch-data, so I need to

install it from usb-modeswitch-data.

Frankly, I am not actually sure if usb_modeswitch_dispatcher is really

needed for the USB_ModeSwitch service type. In much earlier (2010)

versions of USB_ModeSwitch

(https://web.archive.org/web/20100921062014/http://draisberghof.de/usb_modeswitch/usb-modeswitch-data-20100826.tar.bz2)

the UDEV rules file referred to usb_modeswitch_dispatcher. It does

not do so anymore, instead it apparently uses usb_modeswitch directly.

I will ask in the USB_ModeSwitch forums tomorrow if the dispatcher is

needed. I will also ask about an option for specifying the config

file. usb_modeswitch has one, usb_modeswitch_dispatcher does not have

the option.

Regards,

Florian

Find attached patches that make USB modems just work, fixing this bug.

I have confirmed that at least on x86_64-linux

(service usb-modeswitch-service-type)

works with the default config and that

(service usb-modeswitch-service-type

(usb-modeswitch-configuration

(config-file #f)))

works, passing no config file, leading to different defaults. Manual

use as described in a comment of the usb-modeswitch package works as

well.

I hope using putenv before wrap-program is OK. I tried using

computed-file #:options instead, but could not get it to work.

I am uncertain if having paragraphs like this in

modem-manager-service-type and usb-modeswitch-service-type is a good

idea:

This service is part of ‘%desktop-services’ (*note Desktop

Services::).

On Sat, May 25, 2019 at 09:42:15PM +0200, pelzflorian (Florian Pelz) wrote:

I am sure, it is needed.

There are two different executable files called usb_modeswitch. The

one used by the UDEV rule runs the dispatcher.

I have not asked. I will propose the patch after this is commited at

Guix.

Regards,

Florian

From dc50f979f77d9298e8c81e6943076c612e5f459a Mon Sep 17 00:00:00 2001

Fixes https://bugs.gnu.org/35640.

---

doc/guix.texi | 3 +++

gnu/services/desktop.scm | 1 +

2 files changed, 4 insertions(+)

On Wed, Jun 12, 2019 at 12:36:47PM +0200, pelzflorian (Florian Pelz) wrote:

I forgot to ask: The USB_ModeSwitch website says:

If you are a system integrator (package maintainer) you can use this

XML file to check for new releases: usb-modeswitch-versions.xml

http://www.draisberghof.de/usb_modeswitch/usb-modeswitch-versions.xml.

Is it possible to add a hash of this URL to the usb-modeswitch and

usb-modeswitch-data packages so that guix refresh notifies us of new

releases?

On Sat, May 25, 2019 at 07:44:37PM +0200, Ludovic Courtès wrote:

I see now, you mean the exclamation mark? It is not needed. Sorry.

Attached is the same patch without exclamation mark.

Regards,

Florian

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

I meant that the ‘patch-shebangs’ phase should take care of this, so you

can remove this substitution and save two lines.

Ludo’.

On Wed, Jun 12, 2019 at 02:21:05PM +0200, Ludovic Courtès wrote:

It is needed. Upstream USB_Modeswitch bundles jimtcl. Its Makefile

uses jimsh0 from the bundled jimtcl, if possible, and replaces the

shebang using sed. Therefore it seems right to use jimsh for Guix’

USB_ModeSwitch as well.

patch-shebang fails because the basename of #!/usr/bin/tclsh is not

jimsh.

patch-shebang: /gnu/store/24yzny7szv1fnmq6ci1x2ndyy7dzyigk-usb-modeswitch-2.5.2-dispatcher/bin/usb_modeswitch_dispatcher: warning: no binary for interpreter `tclsh' found in $PATH

Regards,

Florian

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

Oh got it, thanks for explaining.

Ludo’.

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

Applied, thanks!

Ludo'.

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

No, it’s not that simple. You could write an “updater” (see (guix

upstream)) that specifically reads this file and reports the new version

specified in this file, but maybe it’s overkill to do that for a single

package.

Thanks,

Ludo’.

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

I think it’s OK.

Looks like we’re almost there! A couple of comments and suggestions:

[...]

s/UDEV/udev/ (here and elsewhere).

(computed-file …) should probably be moved to a separate procedure, to

keep the main procedure concise.

Instead of (case …), maybe this would be slightly clearer:

#$(if config-file

#~(string-append … #$config-file)

"")

Perhaps add comments to explain what’s happening here.

Why do you think so? ‘wrap-program’ does not read $PATH, AFAICS.

OK!

Thank you!

Ludo’.

Find attached patches taking into account your suggestions. (The

second patch is unchanged.)

On Wed, Jun 12, 2019 at 11:11:16PM +0200, Ludovic Courtès wrote:

wrap-program uses (which "bash"). Without bash in the path, it

generates a wrong shebang “#!#f”.

Regards,

Florian

On Thu, Jun 13, 2019 at 08:04:52PM +0200, pelzflorian (Florian Pelz) wrote:

Forgot attachment. Sorry.

From f4a5103b365f8bc5d2c514dad2e40f3c9981cfc2 Mon Sep 17 00:00:00 2001

Fixes https://bugs.gnu.org/35640.

---

doc/guix.texi | 3 +++

gnu/services/desktop.scm | 1 +

2 files changed, 4 insertions(+)

Hello!

"pelzflorian (Florian Pelz)" <pelzflorian@pelzflorian.de> skribis:

[...]

Pushed, thank you!

Ludo’.