This is the mail archive of the
kawa@sourceware.org
mailing list for the Kawa project.
Re: Kawa 1.13 compilation problem, possible regression
- From: Matthieu Vachon <matthieu dot o dot vachon at gmail dot com>
- To: kawa at sourceware dot org
- Date: Fri, 7 Jun 2013 10:37:24 -0400
- Subject: Re: Kawa 1.13 compilation problem, possible regression
- References: <CAOTvmom6rqn9_fgDTjOr=Hfe-jLsvso8ii5KgnT2CgTYYmo0XQ at mail dot gmail dot com> <511026D7 dot 5030007 at bothner dot com> <51551844 dot 5080201 at bothner dot com> <CAOTvmoknAS5EQc0JpemSPnLSRKoPnWgMBQbe5ovbu=n74bZSuw at mail dot gmail dot com> <CAOTvmomjQAq8uOa3UD2Yzs+n+P4C5bpiT9yDFY0stepueWYKpQ at mail dot gmail dot com> <51860D80 dot 6040406 at bothner dot com> <CAOTvmompbGyXXGNaTNhnO3wpdkkb=3CO75Sct-ZUUdNbBh1Eww at mail dot gmail dot com> <51880DE4 dot 5030204 at bothner dot com> <CAOTvmonDEnr4Lnf5i45BpZd6NHE4ev0rGVyGHRKVsbfDyp9Aug at mail dot gmail dot com> <CAOTvmo=xUqLE+MCD441YaodmVodPZe-f3aAeB0urPh=VUNy0vw at mail dot gmail dot com>
By removing the `IfLet` rewrite optimization, I was able to compile my
whole module correctly with Kawa latest version. So it seems that the
optimization make the bug #39150 to show up.
If I find time this weekend, I will try to find what could cause bug
#39151 to happen.
On Mon, Jun 3, 2013 at 10:11 AM, Matthieu Vachon
<matthieu.o.vachon@gmail.com> wrote:
> Ok, created two bug reports on this:
>
> 1. http://savannah.gnu.org/bugs/?39150
> 2. http://savannah.gnu.org/bugs/?39151
>
> Matt
>
> On Mon, Jun 3, 2013 at 9:44 AM, Matthieu Vachon
> <matthieu.o.vachon@gmail.com> wrote:
>> Hi Per,
>>
>> I worked on this possible regression this weekend. From my findings
>> there could be two distinct issues in my file. Finally, it does not
>> have nothing to do with the fact the I'm calling a function from
>> within an object's method which creates an instance of the said
>> object.
>>
>> I created two simple test cases that I used to reproduce the errors.
>> Fixing one could fix the other, but I'm not sure. For each test cases,
>> I isolated the exact changeset that breaks the test. For one of them,
>> I dig up further to try to fix it but was not able to.
>>
>> I will fill two bug reports about those issues.
>>
>> Regards,
>> Matt
>>
>> On Mon, May 6, 2013 at 4:09 PM, Per Bothner <per@bothner.com> wrote:
>>> On 05/05/2013 03:36 PM, Matthieu Vachon wrote:
>>>
>>>> The attached patch did the job. I was able to compile Kawa latest
>>>> version with java5.
>>>
>>>
>>> Ok - I checked it in.
>>>
>>>
>>>> But I faced another problem related with compilation and it could be a
>>>> regression. Where my code compiled fine on 1.12, it was not able to
>>>> compile with latest version. After some some research, it may be
>>>> caused by the way the code is constructed. In the module I'm trying to
>>>> compile (module.scm), we define a class via `define-simple-class` (let
>>>> say its called <Simple>) and inside one of its method, we call a
>>>> procedure defined in the same module.
>>>>
>>>> In this procedure, we create an instance of class <Simple> and use to
>>>> compute a result which is returned. When commenting the line calling
>>>> the procedure in the method, the compilation pass. Maybe it pass as a
>>>> side effects just because some analysis did not occur because the
>>>> procedure call was commented out.
>>>>
>>>> I tried to reproduce the problem with a stripped down version of the
>>>> module but I was not able to do so. I will try to upgrade my test file
>>>> to see if I can reproduce the problem in the upcoming week. I attached
>>>> the file so my explanations are easier to follow.
>>>
>>>
>>> Nothing obvious comes to mind. But it is certainly possible the
>>> compiler is confused about whether an execution path can happen.
>>>
>>> I'll see if you can find a stripped down example.
>>>
>>> --
>>> --Per Bothner
>>> per@bothner.com http://per.bothner.com/