headers
Ivan Maidanski | 28 May 2012 09:40
Picon

Re[12]: [Gc] There should be a library major bump of gc library

Hi Paulo,

Mon, 28 May 2012 02:22:13 -0300 от Paulo César Pereira de Andrade<paulo.cesar.pereira.de.andrade@...>:
> 2012/5/27 Ivan Maidanski <ivmai@...>:
> > Hi Juan and Paulo,
> >
> > Paulo -
> > I've quickly glanced into the commits. I think it is not a good (or even disallowed) thing to alter 7.2 API
(unlike v7.3+).
> > The only commits to release branch now are the ones fixing some bugs or regression (hiding some symbols
actively used by 3rd-party well-know projects could be considered as a regression). (The only reason for
making such fixing should be making legacy code work as expected nothing more.)
> 
>   Fedora 16 has 7.2alpha, and same for Fedora 17 beta, so it will
> have all symbols exported there (ecl works), but rawhide was
> update to 7.2 release 14 May.

Exactly. We need to fix this regression in 7.2.

> 
> > I guess Hans is having the similar opinion of it.
> >
> > So, part of your patch is definitely ok for master branch. But no code of it suitable for release branch for now.
> > Thank you for the quick response but please adjust it (being more close to the fix scenario I've outlined
in my previous post).
> 
>   Ok. The commit in master adds the GC_push* required by ecl.
> It should be ok to make a "request for enhancement" in Fedora
> rawhide to update to gc-7.3alpha if the commit to master is good.
(Continue reading)

Permalink | Reply |
headers
Paulo César Pereira de Andrade | 28 May 2012 16:09
Picon

Re: Re[12]: There should be a library major bump of gc library

2012/5/28 Ivan Maidanski <ivmai@...>:
> Hi Paulo,
>
> Mon, 28 May 2012 02:22:13 -0300 от Paulo César Pereira de Andrade<paulo.cesar.pereira.de.andrade@...>:
>> 2012/5/27 Ivan Maidanski <ivmai@...>:
>> > Hi Juan and Paulo,
>> >
>> > Paulo -
>> > I've quickly glanced into the commits. I think it is not a good (or even disallowed) thing to alter 7.2 API
(unlike v7.3+).
>> > The only commits to release branch now are the ones fixing some bugs or regression (hiding some symbols
actively used by 3rd-party well-know projects could be considered as a regression). (The only reason for
making such fixing should be making legacy code work as expected nothing more.)
>>
>>   Fedora 16 has 7.2alpha, and same for Fedora 17 beta, so it will
>> have all symbols exported there (ecl works), but rawhide was
>> update to 7.2 release 14 May.
>
> Exactly. We need to fix this regression in 7.2.
>
>>
>> > I guess Hans is having the similar opinion of it.
>> >
>> > So, part of your patch is definitely ok for master branch. But no code of it suitable for release branch
for now.
>> > Thank you for the quick response but please adjust it (being more close to the fix scenario I've outlined
in my previous post).
>>
>>   Ok. The commit in master adds the GC_push* required by ecl.
>> It should be ok to make a "request for enhancement" in Fedora
(Continue reading)

Permalink | Reply |
headers
Ivan Maidanski | 2 Jun 2012 19:11
Picon

Re[14]: [Gc] There should be a library major bump of gc library

Hi Paulo and Juan,

Sorry for the delay.

I decided to prepare the required fix myself.
Proposed changes for 7.2 ("gc-7_2-hotfix-2" branch):
https://github.com/ivmai/bdwgc/compare/release...4be945a8 (2 commits, public headers are unchanged)
For 7.3alpha3 ("master" branch): https://github.com/ivmai/bdwgc/compare/f85b8bf6bd...master (3
commits, one of which is the same as for release branch, 4 functions added to GC API)

Please test both branches. If the fix solves the problem then I'm sure we'll release 7.2c tarball soon.

Regards,
Ivan

Mon, 28 May 2012 11:09:21 -0300 от Paulo César Pereira de Andrade<paulo.cesar.pereira.de.andrade@...>:
> 2012/5/28 Ivan Maidanski <ivmai@...>:
> > Hi Paulo,
> >
> > Mon, 28 May 2012 02:22:13 -0300 от Paulo César Pereira de
> Andrade<paulo.cesar.pereira.de.andrade@...>:
> >> 2012/5/27 Ivan Maidanski <ivmai@...>:
> >> > Hi Juan and Paulo,
> >> >
> >> > Paulo -
> >> > I've quickly glanced into the commits. I think it is not a good (or even
> disallowed) thing to alter 7.2 API (unlike v7.3+).
> >> > The only commits to release branch now are the ones fixing some bugs or
> regression (hiding some symbols actively used by 3rd-party well-know projects
> could be considered as a regression). (The only reason for making such fixing
(Continue reading)

Permalink | Reply |
headers
Paulo César Pereira de Andrade | 2 Jun 2012 20:18
Picon

Re: Re[14]: There should be a library major bump of gc library

2012/6/2 Ivan Maidanski <ivmai@...>:
> Hi Paulo and Juan,

  Hi Ivan,

> Sorry for the delay.

  No problems :-)

> I decided to prepare the required fix myself.
> Proposed changes for 7.2 ("gc-7_2-hotfix-2" branch):
https://github.com/ivmai/bdwgc/compare/release...4be945a8 (2 commits, public headers are unchanged)
> For 7.3alpha3 ("master" branch): https://github.com/ivmai/bdwgc/compare/f85b8bf6bd...master (3
commits, one of which is the same as for release branch, 4 functions added to GC API)

  I think it is a far better compromise then my patch to declare the functions
in the "release" branch, and just not hide them. I applied a diff to 7.2b
(it was updated to 7.2b in fedora rawhide a few days ago), rebuilt fedora rpm,
installed it, then built ecl, and it did build everything correctly,
that is, it did
load, compile lisp code, etc.

> Please test both branches. If the fix solves the problem then I'm sure we'll release 7.2c tarball soon.

  I also tested the build from branch master, and it corrects the ecl issues,
and works in a full ecl build.

  I am waiting for gc-7.2c :-)

> Regards,
(Continue reading)

Permalink | Reply |
headers
Juan Jose Garcia-Ripoll | 2 Jun 2012 21:09

Re: Re[14]: There should be a library major bump of gc library

I would like to thank you both, Paulo and Ivan, for solving this problem.

It is really appreciated

Juanjo

--
Instituto de Física Fundamental, CSIC
c/ Serrano, 113b, Madrid 28006 (Spain)
http://juanjose.garciaripoll.googlepages.com
_______________________________________________
Gc mailing list
Gc@...
http://www.hpl.hp.com/hosted/linux/mail-archives/gc/
Permalink | Reply |
headers
Ivan Maidanski | 5 Jun 2012 11:07
Picon

Re[16]: [Gc] There should be a library major bump of gc library

Hi Paulo, Juan and Tsugutomo,

Could you please check the code that is 7.2c candidate (before we make a tarball) and confirm that it works
and solves the problems mentioned by you: https://github.com/ivmai/bdwgc/tree/gc-7_2-hotfix-2

Thank you

Regards,
Ivan

Sat, 2 Jun 2012 14:18:54 -0400 Paulo César Pereira de Andrade <paulo.cesar.pereira.de.andrade@...>:
> 2012/6/2 Ivan Maidanski <ivmai@...>:
> > Hi Paulo and Juan,
> 
>   Hi Ivan,
> 
> > Sorry for the delay.
> 
>   No problems :-)
> 
> > I decided to prepare the required fix myself.
> > Proposed changes for 7.2 ("gc-7_2-hotfix-2" branch):
https://github.com/ivmai/bdwgc/compare/release...4be945a8 (2 commits, public headers are unchanged)
> > For 7.3alpha3 ("master" branch): https://github.com/ivmai/bdwgc/compare/f85b8bf6bd...master
(3 commits, one of which is the same as for release branch, 4 functions added to GC API)
> 
>   I think it is a far better compromise then my patch to declare the functions
> in the "release" branch, and just not hide them. I applied a diff to 7.2b
> (it was updated to 7.2b in fedora rawhide a few days ago), rebuilt fedora rpm,
> installed it, then built ecl, and it did build everything correctly,
> that is, it did
> load, compile lisp code, etc.
> 
> > Please test both branches. If the fix solves the problem then I'm sure we'll release 7.2c tarball soon.
> 
>   I also tested the build from branch master, and it corrects the ecl issues,
> and works in a full ecl build.
> 
>   I am waiting for gc-7.2c :-)
> 
> > Regards,
> > Ivan
> 
> Paulo
>
Permalink | Reply |
headers
tsugutomo.enami | 5 Jun 2012 12:23
Picon

Re: There should be a library major bump of gc library

Ivan Maidanski <ivmai@...> writes:

> Could you please check the code that is 7.2c candidate (before we make
> a tarball) and confirm that it works and solves the problems mentioned
> by you: https://github.com/ivmai/bdwgc/tree/gc-7_2-hotfix-2

Hi.

I've tested it (make and make check) on both NetBSD/i386 6.0_BETA2 and
NetBSD/amd64 6.0_BETA2.  The output of last several lines from the
latter is below.

enami.

PASS: threadleaktest
PASS: threadkey_test
subthread_create: created 195 threads (193 ended)
PASS: subthread_create
PASS: initsecondarythread
Threaded disclaim test.
PASS: disclaim_test
                        fin. ratio       time/s    time/fin.
regular finalization:       0.9922         4.99  1.19908e-06
 finalize on reclaim:       1.0000         1.36  3.24249e-07
     no finalization:       0.0000         0.27          N/A
PASS: disclaim_bench
===================
All 14 tests passed
===================
Permalink | Reply |
headers
Ivan Maidanski | 7 Jun 2012 23:57
Picon

Re[2]: [Gc] There should be a library major bump of gc library

Hi,

I've updated https://github.com/ivmai/bdwgc/tree/gc-7_2-hotfix-2 branch by adding a fix preventing
null pointer dereference in a rare case of out-of-memory (this bug has been recently discovered by Dmitry
Kuteynikov using a static analysis tool).

Patch applied: https://github.com/ivmai/bdwgc/commit/0ac7a655aba6c7e1d43ad2262e4c556879e89827

Regards,
Ivan

Tue, 05 Jun 2012 19:23:44 +0900 tsugutomo.enami@...:
> Ivan Maidanski <ivmai@...> writes:
> 
> > Could you please check the code that is 7.2c candidate (before we make
> > a tarball) and confirm that it works and solves the problems mentioned
> > by you: https://github.com/ivmai/bdwgc/tree/gc-7_2-hotfix-2
> 
> Hi.
> 
> I've tested it (make and make check) on both NetBSD/i386 6.0_BETA2 and
> NetBSD/amd64 6.0_BETA2.  The output of last several lines from the
> latter is below.
> 
> enami.
> 
> PASS: threadleaktest
> PASS: threadkey_test
> subthread_create: created 195 threads (193 ended)
> PASS: subthread_create
> PASS: initsecondarythread
> Threaded disclaim test.
> PASS: disclaim_test
>                         fin. ratio       time/s    time/fin.
> regular finalization:       0.9922         4.99  1.19908e-06
>  finalize on reclaim:       1.0000         1.36  3.24249e-07
>      no finalization:       0.0000         0.27          N/A
> PASS: disclaim_bench
> ===================
> All 14 tests passed
> ===================
> _______________________________________________
> Gc mailing list
> Gc@...
> http://www.hpl.hp.com/hosted/linux/mail-archives/gc/
>
Permalink | Reply |

Gmane