GNU bug report logs - #29226
fresh-auto-compile doesn’t invalidate the compilation cache

Previous Next

Full log

View this message in rfc822 format

From: help-debbugs <at> gnu.org (GNU bug Tracking System)
To: ludo <at> gnu.org (Ludovic Courtès)
Cc: tracker <at> debbugs.gnu.org
Subject: bug#29226: closed (fresh-auto-compile doesn’t
 invalidate the compilation cache)
Date: Mon, 15 Jan 2018 22:22:02 +0000

[Message part 1 (text/plain, inline)]

Your message dated Mon, 15 Jan 2018 23:21:19 +0100
with message-id <87h8rmg3vk.fsf <at> gnu.org>
and subject line Re: bug#29226: fresh-auto-compile doesn’t invalidate the compilation cache
has caused the debbugs.gnu.org bug report #29226,
regarding fresh-auto-compile doesn’t invalidate the compilation cache
to be marked as done.

(If you believe you have received this mail in error, please contact
help-debbugs <at> gnu.org.)


-- 
29226: http://debbugs.gnu.org/cgi/bugreport.cgi?bug=29226
GNU Bug Tracking System
Contact help-debbugs <at> gnu.org with problems

[Message part 2 (message/rfc822, inline)]

From: ludovic.courtes <at> inria.fr (Ludovic Courtès)
To: bug-guile <at> gnu.org
Subject: fresh-auto-compile doesn’t invalidate the
 compilation cache
Date: Thu, 09 Nov 2017 13:53:22 +0100

[Message part 3 (text/plain, inline)]

Hello,

‘%fresh-auto-compile’ is documented as a way to “forcibly invalidate the
auto-compilation cache” (info "(guile) Compilation"), but it seems that
it doesn’t invalidate much.  Specifically, ‘primitive-load-path’ does:

  compiled_thunk = load_thunk_from_path (filename, full_filename, &stat_source,
                                         &found_stale_compiled_file);

  if (scm_is_false (compiled_thunk)
      && scm_is_false (*scm_loc_fresh_auto_compile)
      …)
    { … }

  …

  if (scm_is_true (compiled_thunk))
    return scm_call_0 (compiled_thunk);
  else
    {
      SCM freshly_compiled = scm_try_auto_compile (full_filename);
      …
    }

So if there’s a .go file in the search path, it is *always* loaded, and
there’s no way we reach ‘scm_try_auto_compile’.

‘load-absolute’ in boot-9.scm seems to have the same problem:

  (and scmstat (or (pre-compiled) (fallback)))

I believe the attached patch fixes it.

Thoughts?

Thanks,
Ludo’.

[Message part 4 (text/x-patch, inline)]

diff --git a/libguile/load.c b/libguile/load.c
index 7b8136af8..3747dd9ad 100644
--- a/libguile/load.c
+++ b/libguile/load.c
@@ -1,5 +1,5 @@
 /* Copyright (C) 1995, 1996, 1998, 1999, 2000, 2001, 2004, 2006, 2008,
- *   2009, 2010, 2011, 2012, 2013, 2014 Free Software Foundation, Inc.
+ *   2009, 2010, 2011, 2012, 2013, 2014, 2017 Free Software Foundation, Inc.
  * 
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public License
@@ -1195,13 +1195,16 @@ SCM_DEFINE (scm_primitive_load_path, "primitive-load-path", 0, 0, 1,
                                *scm_loc_load_extensions, SCM_BOOL_F,
                                &stat_source);
 
-  compiled_thunk = load_thunk_from_path (filename, full_filename, &stat_source,
-                                         &found_stale_compiled_file);
+  if (scm_is_false (*scm_loc_fresh_auto_compile))
+    compiled_thunk = load_thunk_from_path (filename, full_filename,
+					   &stat_source,
+					   &found_stale_compiled_file);
+  else
+    compiled_thunk = SCM_BOOL_F;
 
   if (scm_is_false (compiled_thunk)
       && scm_is_true (full_filename)
       && scm_is_true (*scm_loc_compile_fallback_path)
-      && scm_is_false (*scm_loc_fresh_auto_compile)
       && scm_is_pair (*scm_loc_load_compiled_extensions)
       && scm_is_string (scm_car (*scm_loc_load_compiled_extensions)))
     {
diff --git a/module/ice-9/boot-9.scm b/module/ice-9/boot-9.scm
index 7127d73f7..612040aa4 100644
--- a/module/ice-9/boot-9.scm
+++ b/module/ice-9/boot-9.scm
@@ -3812,7 +3812,10 @@ when none is available, reading FILE-NAME with READER."
                                               scmstat
                                               go-file-name))))))
 
-    (let ((compiled (and scmstat (or (pre-compiled) (fallback)))))
+    (let ((compiled (and scmstat
+                         (or (and (not %fresh-auto-compile)
+                                  (pre-compiled))
+                             (fallback)))))
       (if compiled
           (begin
             (if %load-hook

[Message part 5 (message/rfc822, inline)]

From: ludo <at> gnu.org (Ludovic Courtès)
To: Andy Wingo <wingo <at> igalia.com>
Cc: 29226-done <at> debbugs.gnu.org
Subject: Re: bug#29226: fresh-auto-compile doesn’t
 invalidate the compilation cache
Date: Mon, 15 Jan 2018 23:21:19 +0100

Andy Wingo <wingo <at> igalia.com> skribis:

> On Mon 15 Jan 2018 15:33, ludo <at> gnu.org (Ludovic Courtès) writes:
>
>> Hello,
>>
>> ludovic.courtes <at> inria.fr (Ludovic Courtès) skribis:
>>
>>> ‘%fresh-auto-compile’ is documented as a way to “forcibly invalidate the
>>> auto-compilation cache” (info "(guile) Compilation"), but it seems that
>>> it doesn’t invalidate much.  Specifically, ‘primitive-load-path’ does:
>>>
>>>   compiled_thunk = load_thunk_from_path (filename, full_filename, &stat_source,
>>>                                          &found_stale_compiled_file);
>>>
>>>   if (scm_is_false (compiled_thunk)
>>>       && scm_is_false (*scm_loc_fresh_auto_compile)
>>>       …)
>>>     { … }
>>>
>>>   …
>>>
>>>   if (scm_is_true (compiled_thunk))
>>>     return scm_call_0 (compiled_thunk);
>>>   else
>>>     {
>>>       SCM freshly_compiled = scm_try_auto_compile (full_filename);
>>>       …
>>>     }
>>>
>>> So if there’s a .go file in the search path, it is *always* loaded, and
>>> there’s no way we reach ‘scm_try_auto_compile’.
>>>
>>> ‘load-absolute’ in boot-9.scm seems to have the same problem:
>>>
>>>   (and scmstat (or (pre-compiled) (fallback)))
>>>
>>> I believe the attached patch fixes it.
>>
>>> Pushed as 83d4c4d622b406ec0bc9d8139ec8182fa72b5720.
>>
>> As we’ve discussed since then, this patch was misguided:
>> ‘%fresh-auto-compile’ is about ~/.cache, so the fact that .go files
>> found elsewhere in the search path get loaded is expected.
>>
>> My apologies for breaking this!
>>
>> So I guess we should revert 83d4c4d622b406ec0bc9d8139ec8182fa72b5720 and
>> release 2.2.4 afterwards.  Does that make sense, Andy?
>
> Makes sense to me, yes!

OK, reverted in a1793bc904d947810d64175945218c2cdace38cc.

Thanks,
Ludo’.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.