[asdf-devel] Still having problems on upgrade tests

[Robert Goldman]

On 3/3/13 Mar 3 -8:08 PM, Faré wrote:
> I don't understand what could be going on. Of course, and especially
> so when we're testing upgrades, there's plenty of pathname magic and
> configuration switching going on. But I can't imagine what's at stake
> to make it work for me and not for you. Are you using the latest
> checkout from the master branch, as opposed to the release branch?
> 
> Did you try to make mrproper and/or git clean -xfd to remove any
> parasite files from your checkout?

Yes, and I did a git diff to check.

The following seems to be the crux of the issue:

; Upgraded ASDF from version 1.85 to version 2.31.8
; Registering #<system "asdf">
;;; Writing fasl file /Users/rpg/lisp/asdf/build/asdf.fasl

this should be in an implementation-specific subdirectory, but isn't.

Is there any chance this could be because I am not running bash?  Or
somehow something is briefly turning off the output-redirection?

I'll try to pry into this tomorrow.

best,
r

> 
> —♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
> Lubarsky's Law of Cybernetic Entomology:
>         There's always one more bug.
> 
> 
> On Sun, Mar 3, 2013 at 8:46 PM, Robert Goldman <rpgoldman at sift.info> wrote:
>> On 3/3/13 Mar 3 -5:38 PM, Faré wrote:
>>> On Sun, Mar 3, 2013 at 5:41 PM, Robert Goldman <rpgoldman at sift.info> wrote:
>>>> The upgrade test for ACL from 1.85 fails reliably with this error:
>>>>
>>>> Warning: COMPILE-FILE warned while performing #<compile-op > on
>>>> #<cl-source-file "asdf" "build" "asdf">.
>>>> Warning: COMPILE-FILE failed while performing #<compile-op > on
>>>> #<cl-source-file "asdf" "build" "asdf">.
>>>> TEST ABORTED:
>>>> #P"/Users/rpg/lisp/asdf/build/fasls/acl-8.2m-macosx-x64/asdf/build/asdf.fasl"
>>>> does not exist, cannot load
>>> [...]
>>>> Script failed
>>>> upgrade FAILED for allegromodern from 1.85 using method
>>>> 'load-asdf-lisp'load-asdf-system
>>>>
>>>> Interestingly, when I paste the replication string into bash:
>>>>
>>>> ASDF_UPGRADE_TEST_TAGS="1.85"
>>>> ASDF_UPGRADE_TEST_METHODS="'load-asdf-lisp'load-asdf-system"
>>>> ./test/run-tests.sh -u allegromodern
>>>>
>>>> this works fine.
>>>>
>>>> So this only fails for me when running in the context of make....
>>>>
>>> Works for me, at least with Allegro 9.0:
>>>
>>> make u l=allegro ASDF_UPGRADE_TEST_TAGS=1.85
>>> make u l=allegromodern ASDF_UPGRADE_TEST_TAGS=1.85
>>
>> I'm stumped.  It fails for me on Allegro 9.0 just as with 8.2
>>
>> Note that this happens for me in the context of 'make test-all'
>>
>>
>> make u l=allegromodern ASDF_UPGRADE_TEST_TAGS=1.85
>>
>> also fails...
>>
>> Somehow the build is not working:
>>
>> rpg% ls
>> /Users/rpg/lisp/asdf/build/fasls/acl-9.0m-macosx-x64/asdf/build/asdf.fasl
>> ls: cannot access
>> /Users/rpg/lisp/asdf/build/fasls/acl-9.0m-macosx-x64/asdf/build/asdf.fasl:
>> No such file or directory
>>
>> For some reason, I have no asdf.fasl there, but I *do* have an asdf.lisp...
>>
>> I see this, which indicates that the fasl is being written in the wrong
>> location:
>>
>> ; Registering #<system "asdf">
>> ;;; Writing fasl file /Users/rpg/lisp/asdf/build/asdf.fasl
>> ;;; Fasl write complete
>> Warning: COMPILE-FILE warned while performing #<compile-op > on
>> #<cl-source-file "asdf" "build" "asdf">.
>> Warning: COMPILE-FILE failed while performing #<compile-op > on
>> #<cl-source-file "asdf" "build" "asdf">.
>>
>> And that's the file, alright:
>>
>> pg% head build/asdf.fasl
>> ?z??#<<AcL>> /Users/rpg/lisp/asdf/build/asdf.lisp by rpg on
>> rpgoldman-3.local at 2013-03-03T19:40:46+06\
>> using 9.0 [64-bit Mac OS X (Intel)] (Feb 26, 2013 9:53)\
>> fasl version = 63\
>> runtime version = 33\
>> for non-smp lisps; #+8-bit-specific code; #+16-bit-specific code\
>> Optimization settings at wfasl time:\
>> ((safety 3) (space 1) (speed 2) (compilation-speed 1) (debug 2))\
>>
>> So is there something going awry in the build process?
>>
>> Best,
>> r
>>
>>
>>>
>>> I can't try allegro 8.2, because my license has expired, and Franz
>>> only offers one until January 31st 2013, and I don't feel like
>>> cheating on the system date:
>>> http://www.franz.com/products/express/
>>>
>>> Is it a case of confusion whereby we changed the way the
>>> implementation identifier is computed, and asdf creates the fasl in
>>> one directory but somehow looks for it in another?
>>>
>>> What is the command that makes it fail, already?
>>>
>>> —♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org
>>> The kingly office is entitled to no respect. It was originally procured by the
>>> highwayman's methods; it remains a perpetuated crime, can never be anything but
>>> the symbol of a crime. It is no more entitled to respect than is the flag of
>>> a pirate. — Mark Twain
>>>
>>