bug#47031: False positives for ‘-Wunbound-variable’ in 3.0.5

bug#47031: False positives for ‘-Wunbound-variable’ in 3.0.5

[Ludovic Courtès]

The new ‘unbound-variable’ warning in 3.0.5 has false positives, as in
this example:

--8<---------------cut here---------------start------------->8---
$ cat /tmp/unbound.scm
(use-modules (srfi srfi-35))

(condition (&message (message "oh no!"))
           (&serious))
$ guild compile -Wunbound-variable /tmp/unbound.scm
/tmp/unbound.scm:3:0: warning: possibly unbound variable `make-compound-condition'
wrote `/home/ludo/.cache/guile/ccache/3.0-LE-8-4.4/tmp/unbound.scm.go'
$ guild --version
guild (GNU Guile) 3.0.5
Copyright (C) 2021 Free Software Foundation, Inc.

License LGPLv3+: GNU LGPL 3 or later <http://gnu.org/licenses/lgpl.html>.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
--8<---------------cut here---------------end--------------->8---

3.0.4 and earlier versions are fine.

Ludo’.

bug#47031: False positives for ‘-Wunbound-variable’ in 3.0.5

[Ludovic Courtès]

[-- Attachment #1: Type: text/plain, Size: 533 bytes --]

Hello!

Ludovic Courtès <ludo@gnu.org> skribis:

> $ cat /tmp/unbound.scm
> (use-modules (srfi srfi-35))
>
> (condition (&message (message "oh no!"))
>            (&serious))
> $ guild compile -Wunbound-variable /tmp/unbound.scm
> /tmp/unbound.scm:3:0: warning: possibly unbound variable `make-compound-condition'

The problem was that ‘make-compound-condition’ is re-exported from (srfi
srfi-35) and this particular case was not handled.

The attached patch fixes it.  Let me know what you think!

Ludo’.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Type: text/x-patch, Size: 1714 bytes --]

diff --git a/module/language/tree-il/analyze.scm b/module/language/tree-il/analyze.scm
index 766568f38..4b53e426f 100644
--- a/module/language/tree-il/analyze.scm
+++ b/module/language/tree-il/analyze.scm
@@ -510,6 +510,11 @@ given `tree-il' element."
            ;; The variable is an import.  At the time of use, the
            ;; name is bound to the import.
            'import)
+          ((and=> (module-public-interface mod)
+                  (lambda (interface)
+                    (module-variable interface name)))
+           ;; The variable is re-exported from module.
+           'import)
           (else
            ;; Variable unbound in the module.
            'unbound))))))
diff --git a/test-suite/tests/tree-il.test b/test-suite/tests/tree-il.test
index 217a1000f..262c2454d 100644
--- a/test-suite/tests/tree-il.test
+++ b/test-suite/tests/tree-il.test
@@ -557,6 +557,19 @@
                                         #:env m
                                         #:opts %opts-w-unbound))))))))
 
+   (pass-if "macro-inserted binding"
+     (null? (call-with-warnings
+             (lambda ()
+               (compile '(begin
+                           (use-modules (srfi srfi-35))
+
+                           ;; This 'condition' form expands to a
+                           ;; 'make-compound-condition' call, which is
+                           ;; re-exported from (ice-9 exceptions).
+                           (condition (&error)
+                                      (&message (message "oh!"))))
+                        #:opts %opts-w-unbound)))))
+
    (with-test-prefix "use-before-definition"
      (define-syntax-rule (pass-if-warnings expr pat test)
        (pass-if 'expr

bug#47031: False positives for ‘-Wunbound-variable’ in 3.0.5

[Ludovic Courtès]

Ludovic Courtès <ludo@gnu.org> skribis:

> The problem was that ‘make-compound-condition’ is re-exported from (srfi
> srfi-35) and this particular case was not handled.
>
> The attached patch fixes it.  Let me know what you think!

Pushed as 5969490f55e5a167a3eb2573a3c5241c0ce079f3 following the
IRC’d LGTM.  :-)

Ludo’.