Re: search-paths and propagated inputs

Re: search-paths and propagated inputs

[Federico Beffa]

Federico Beffa <beffa@ieee.org> writes:

> Hi,
>
> the package 'gobject-introspection' declares a
> search-path-specification for the variable GI_TYPELIB_PATH and
> 'matplotlib' uses and propagates said package. I have 'matplotlib' in
> my profile, but "guix package --search-paths" doesn't show anything
> about GI_TYPELIB_PATH.
>
> Is it intentional or an oversight that search-path-specifications of
> propagated-inputs are not considered? Or, am I doing something wrong?
>
> Regards,
> Fede

Bug or feature?

Regards,
Fede

Re: search-paths and propagated inputs

[Ricardo Wurmus]

Federico Beffa <beffa@ieee.org> writes:

> Federico Beffa <beffa@ieee.org> writes:
>
>> Hi,
>>
>> the package 'gobject-introspection' declares a
>> search-path-specification for the variable GI_TYPELIB_PATH and
>> 'matplotlib' uses and propagates said package. I have 'matplotlib' in
>> my profile, but "guix package --search-paths" doesn't show anything
>> about GI_TYPELIB_PATH.
>>
>> Is it intentional or an oversight that search-path-specifications of
>> propagated-inputs are not considered? Or, am I doing something wrong?
>>
>> Regards,
>> Fede
>
> Bug or feature?

I think it’s a bug.  Right now users of matplotlib have to check the
sources where it says that GI_TYPELIB_PATH must be set.

~~ Ricardo

bug#22073: Fwd: search-paths and propagated inputs

[Federico Beffa]

Federico Beffa <beffa@ieee.org> writes:

> Federico Beffa <beffa@ieee.org> writes:
>
>> Hi,
>>
>> the package 'gobject-introspection' declares a
>> search-path-specification for the variable GI_TYPELIB_PATH and
>> 'matplotlib' uses and propagates said package. I have 'matplotlib' in
>> my profile, but "guix package --search-paths" doesn't show anything
>> about GI_TYPELIB_PATH.
>>
>> Is it intentional or an oversight that search-path-specifications of
>> propagated-inputs are not considered? Or, am I doing something wrong?
>>
>> Regards,
>> Fede
>
> Bug or feature?

I think it’s a bug.  Right now users of matplotlib have to check the
sources where it says that GI_TYPELIB_PATH must be set.

~~ Ricardo

bug#22073: Fwd: search-paths and propagated inputs

[Federico Beffa]

Looking at my manifest file I see that, e.g., the entry for
'python-matplotlib' lists all recursively propagated
dependencies. Differently from this the 'search-paths' entry only lists
the entries defined in the package (in this case none), neglecting the
search paths in the 'propagated-inputs'.

If I understand correctly, this behavior is defined by this function (from
guix/profiles.scm):

(define* (package->manifest-entry package #:optional output)
  "Return a manifest entry for the OUTPUT of package PACKAGE.  When OUTPUT is
omitted or #f, use the first output of PACKAGE."
  (let ((deps (map (match-lambda
                    ((label package)
                     (gexp-input package))
                    ((label package output)
                     (gexp-input package output)))
                   (package-transitive-propagated-inputs package))))
    (manifest-entry
     (name (package-name package))
     (version (package-version package))
     (output (or output (car (package-outputs package))))
     (item package)
     (dependencies (delete-duplicates deps))
     (search-paths (package-native-search-paths package)))))

To get all the required search paths recursively we should replace the
last 'manifest-entry' slot with a call to a function like this:

(define (collect-package-search-paths package)
  (define (collect-search-paths package)
    (let ((propagated-inputs (package-propagated-inputs package))
          (search-paths (package-native-search-paths package)))
      (append search-paths
              (append-map
               (match-lambda
                ((name pkg . rest) (collect-search-paths pkg))
                (_ '()))
               propagated-inputs))))
  (delete-duplicates (collect-search-paths package)))

Or, am I missing how this works?

Regards,
Fede

bug#22073: Fwd: search-paths and propagated inputs

[Ludovic Courtès]

[-- Attachment #1: Type: text/plain, Size: 2204 bytes --]

Thanks for the heads-up, and sorry for the delay!

Federico Beffa <beffa@ieee.org> skribis:

> Looking at my manifest file I see that, e.g., the entry for
> 'python-matplotlib' lists all recursively propagated
> dependencies. Differently from this the 'search-paths' entry only lists
> the entries defined in the package (in this case none), neglecting the
> search paths in the 'propagated-inputs'.

That’s a bug, indeed.

> If I understand correctly, this behavior is defined by this function (from
> guix/profiles.scm):
>
> (define* (package->manifest-entry package #:optional output)
>   "Return a manifest entry for the OUTPUT of package PACKAGE.  When OUTPUT is
> omitted or #f, use the first output of PACKAGE."
>   (let ((deps (map (match-lambda
>                     ((label package)
>                      (gexp-input package))
>                     ((label package output)
>                      (gexp-input package output)))
>                    (package-transitive-propagated-inputs package))))
>     (manifest-entry
>      (name (package-name package))
>      (version (package-version package))
>      (output (or output (car (package-outputs package))))
>      (item package)
>      (dependencies (delete-duplicates deps))
>      (search-paths (package-native-search-paths package)))))
>
> To get all the required search paths recursively we should replace the
> last 'manifest-entry' slot with a call to a function like this:

Right.

Here’s a variant of what you propose.  With that, I get:

--8<---------------cut here---------------start------------->8---
$ ./pre-inst-env guix package -i python2-matplotlib -p foo

[...]

The following environment variable definitions may be needed:
   export PATH="foo/bin:foo/sbin"
   export PYTHONPATH="foo/lib/python2.7/site-packages"
   export GI_TYPELIB_PATH="foo/lib/girepository-1.0"
   export XDG_DATA_DIRS="foo/share"
   export GIO_EXTRA_MODULES="foo/lib/gio/modules"
--8<---------------cut here---------------end--------------->8---

… which is what we’re expecting, right?

If that’s fine with you, I’ll commit this patch along with test cases.

Thanks!

Ludo’.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Type: text/x-patch, Size: 1771 bytes --]

diff --git a/guix/packages.scm b/guix/packages.scm
index 68fb091..7222337 100644
--- a/guix/packages.scm
+++ b/guix/packages.scm
@@ -89,6 +89,7 @@
             package-transitive-target-inputs
             package-transitive-native-inputs
             package-transitive-propagated-inputs
+            package-transitive-native-search-paths
             package-transitive-supported-systems
             package-source-derivation
             package-derivation
@@ -632,6 +633,16 @@ for the host system (\"native inputs\"), and not target inputs."
 recursively."
   (transitive-inputs (package-propagated-inputs package)))
 
+(define (package-transitive-native-search-paths package)
+  "Return the list of search paths for PACKAGE and its propagated inputs,
+recursively."
+  (append-map (match-lambda
+                ((label (? package? p) _ ...)
+                 (package-native-search-paths p))
+                (_
+                 '()))
+              (package-transitive-propagated-inputs package)))
+
 (define (transitive-input-references alist inputs)
   "Return a list of (assoc-ref ALIST <label>) for each (<label> <package> . _)
 in INPUTS and their transitive propagated inputs."
diff --git a/guix/profiles.scm b/guix/profiles.scm
index ce6b2c4..ce86ff8 100644
--- a/guix/profiles.scm
+++ b/guix/profiles.scm
@@ -176,7 +176,7 @@ omitted or #f, use the first output of PACKAGE."
      (output (or output (car (package-outputs package))))
      (item package)
      (dependencies (delete-duplicates deps))
-     (search-paths (package-native-search-paths package)))))
+     (search-paths (package-transitive-native-search-paths package)))))
 
 (define (packages->manifest packages)
   "Return a list of manifest entries, one for each item listed in PACKAGES.

bug#22073: Fwd: search-paths and propagated inputs

[Federico Beffa]

On Sun, Dec 20, 2015 at 1:56 PM, Ludovic Courtès <ludo@gnu.org> wrote:
>> To get all the required search paths recursively we should replace the
>> last 'manifest-entry' slot with a call to a function like this:
>
> Right.
>
> Here’s a variant of what you propose.  With that, I get:
>
> --8<---------------cut here---------------start------------->8---
> $ ./pre-inst-env guix package -i python2-matplotlib -p foo
>
> [...]
>
> The following environment variable definitions may be needed:
>    export PATH="foo/bin:foo/sbin"
>    export PYTHONPATH="foo/lib/python2.7/site-packages"
>    export GI_TYPELIB_PATH="foo/lib/girepository-1.0"
>    export XDG_DATA_DIRS="foo/share"
>    export GIO_EXTRA_MODULES="foo/lib/gio/modules"
> --8<---------------cut here---------------end--------------->8---
>
> … which is what we’re expecting, right?

I think so.

>
> If that’s fine with you, I’ll commit this patch along with test cases.

Sure, thanks for that!

Regards,
Fede

bug#22073: Fwd: search-paths and propagated inputs

[Ludovic Courtès]

Fixed in ccda8f7, thanks!

Ludo’.

bug#22073: Fwd: search-paths and propagated inputs

[Federico Beffa]

Any feedback on this?

Thanks,
Fede

Re: search-paths and propagated inputs

[Ludovic Courtès]

Federico Beffa <beffa@ieee.org> skribis:

> Federico Beffa <beffa@ieee.org> writes:
>
>> Hi,
>>
>> the package 'gobject-introspection' declares a
>> search-path-specification for the variable GI_TYPELIB_PATH and
>> 'matplotlib' uses and propagates said package. I have 'matplotlib' in
>> my profile, but "guix package --search-paths" doesn't show anything
>> about GI_TYPELIB_PATH.
>>
>> Is it intentional or an oversight that search-path-specifications of
>> propagated-inputs are not considered? Or, am I doing something wrong?
>>
>> Regards,
>> Fede
>
> Bug or feature?

Sorry to reply to a months-old message…  and thanks for waiting.  :-)

What you describe above would be a bug, but it seems there’s no longer
any such bug:

--8<---------------cut here---------------start------------->8---
scheme@(guile-user)> ,m (gnu packages python)
scheme@(gnu packages python)> (package-search-paths python-matplotlib)
$2 = ()
scheme@(gnu packages python)> (package-transitive-native-search-paths python-matplotlib)
$3 = (#<<search-path-specification> variable: "GI_TYPELIB_PATH" files: ("lib/girepository-1.0") separator: ":" file-type: directory file-pattern: #f> #<<search-path-specification> variable: "PYTHONPATH" files: ("lib/python3.4/site-packages") separator: ":" file-type: directory file-pattern: #f> #<<search-path-specification> variable: "GUIX_GTK3_PATH" files: ("lib/gtk-3.0") separator: ":" file-type: directory file-pattern: #f> #<<search-path-specification> variable: "XDG_DATA_DIRS" files: ("share") separator: ":" file-type: directory file-pattern: #f> #<<search-path-specification> variable: "GIO_EXTRA_MODULES" files: ("lib/gio/modules") separator: ":" file-type: directory file-pattern: #f>)
--8<---------------cut here---------------end--------------->8---

Could you confirm?

Ludo’.

Re: search-paths and propagated inputs

[Federico Beffa]

On Mon, Mar 7, 2016 at 2:35 PM, Ludovic Courtès <ludo@gnu.org> wrote:

[...]

> Could you confirm?

We fixed that a while back :-)
http://debbugs.gnu.org/cgi/bugreport.cgi?bug=22073

Regards,
Fede

Re: search-paths and propagated inputs

[Ludovic Courtès]

Federico Beffa <beffa@ieee.org> skribis:

> On Mon, Mar 7, 2016 at 2:35 PM, Ludovic Courtès <ludo@gnu.org> wrote:
>
> [...]
>
>> Could you confirm?
>
> We fixed that a while back :-)
> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=22073

Excellent non-news.  :-)

My inbox is a mess!

Cheers,
Ludo’.

search-paths and propagated inputs

[Federico Beffa]

Hi,

the package 'gobject-introspection' declares a
search-path-specification for the variable GI_TYPELIB_PATH and
'matplotlib' uses and propagates said package. I have 'matplotlib' in
my profile, but "guix package --search-paths" doesn't show anything
about GI_TYPELIB_PATH.

Is it intentional or an oversight that search-path-specifications of
propagated-inputs are not considered? Or, am I doing something wrong?

Regards,
Fede