New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Bleadperl between v5.21.7-259-g819b139 and v5.21.7-435-gdf3c16d broke RIBASUSHI/DBIx-Class-TimeStamp-0.14.tar.gz #14528
Comments
From @andkbisect It started to fail with v5.21.7-259-g819b139 but when that one was The brokenness only occurs for perls compiled with -Duselongdouble. sample report http://www.cpantesters.org/cpan/report/49fefdde-a699-11e4-a2b8-12678971dd2f perl -V Summary of my perl5 (revision 5 version 21 subversion 8) configuration: Characteristics of this binary (from libperl): -- |
From @tonycozOn Wed Feb 18 23:20:57 2015, andreas.koenig.7os6VVqR@franz.ak.mind.de wrote:
It was reverted in v5.21.7-436-g13c59d4 not v5.21.7-435-gdf3c16d. I can't reproduce the failure in v5.21.9-22-gf266b74 (blead) or in v5.21.7-436-g13c59d4. I built both with: ./Configure -Dprefix=/home/tony/perl/blead -Uversiononly -Dusedevel -des -Ui_db -Uuseithreads -Duselongdouble -DDEBUGGING=-g Tony |
The RT System itself - Status changed from 'new' to 'open' |
From @andk
tc> On Wed Feb 18 23:20:57 2015, andreas.koenig.7os6VVqR@franz.ak.mind.de wrote:
tc> It was reverted in v5.21.7-436-g13c59d4 not v5.21.7-435-gdf3c16d. Thanks for the corection. tc> I can't reproduce the failure in v5.21.9-22-gf266b74 (blead) or in v5.21.7-436-g13c59d4. Quite surprising for me. tc> I built both with: tc> ./Configure -Dprefix=/home/tony/perl/blead -Uversiononly -Dusedevel -des -Ui_db -Uuseithreads -Duselongdouble -DDEBUGGING=-g Thanks for trying. I cannot explain. I still see the fail with Can you compare you system with http://www.cpantesters.org/cpan/report/cf50d412-a699-11e4-9e0e-9e908971dd2f ? -- |
From @tonycozOn Sun Feb 22 22:17:05 2015, andreas.koenig.7os6VVqR@franz.ak.mind.de wrote:
A newer compiler did the trick. Tony |
From @tonycozOn Mon Feb 23 16:54:54 2015, tonyc wrote:
The problem is occuring in S_outside_integer() which is reporting the bottom of the range in this code: # an attempt to detect former effects of RT#79576, bug itself present between for my $stringifiable++ if ( length ref $bind->[$i][1] and is_plain_value($bind->[$i][1]) ); (the 0) as being out of range because Perl_isinfnan() is returning true. If I revert 415b66b which introduced this check the test passes. If I use the nv value as a memory reference (eg. to dump the raw content of the NV) the tests pass. Just dumping the content of the nv as a number makes no difference. Here's the generated code with some comments: 0000000000000280 <S_outside_integer>: 2fb: 0f 1f 44 00 00 nopl 0x0(%rax,%rax,1) Tony |
From @ribasushiOn 02/25/2015 07:07 AM, Tony Cook via RT wrote:
While this issue has not been resolved, it will likely disappear from Cheers |
From @tonycozOn Fri Mar 20 03:31:17 2015, rabbit-p5p@rabbit.us wrote:
This didn't fix the problem. t/08noclobber.t .... 1/9 Range iterator outside integer range at /home/tony/.cpan/build/DBIx-Class-0.082820-U_G_vt/blib/lib/DBIx/Class/Storage/DBI/SQLite.pm lin $ ~/perl/blead-gcc/bin/perl -MDBIx::Class -E 'say $DBIx::Class::VERSION' Tony |
From @tonycozOn Tue Feb 24 22:07:01 2015, tonyc wrote:
Here's where the strangeness happens: (gdb) Status Word: 0x7069 IE OE PE SF C3 Status Word: 0x6a69 IE OE PE SF C1 C3 The C1 flag being set indicates a stack overflow occurred, so some generated code isn't managing the stack properly. Tony Tony Tony |
From @tonycozOn Wed Apr 08 19:02:25 2015, tonyc wrote:
This is caused by a mismatch between Time::HiRes and Time::Warp on the types of the function pointer stored in the "Time::NVtime" entry of PL_modglobal. Time::HiRes documents that this pointer is: double (*)() but the implementation actually stores a: NV (*)() pointer in that entry. Time::Warp follows the documentation and expects that pointer to be a double(*)() and stores a replacement double(*)() pointer in that entry. On x86_64 double is returned as a value on the FPU stack, but long double is returned the same way as a structure. When the NV version of the function is called through the double pointer, an FPU stack underflow occurs, or accumulates unused values when the reverse happens. In any case, this is undefined behaviour and breaks the tests. I've posted a pull request at manwar/Time-Warp#2 I'll post an issue to Time::HiRes too. This isn't a bug in perl, I'll close this ticket soon. Tony |
From @tonycozOn Wed Apr 15 22:39:20 2015, tonyc wrote:
https://rt.cpan.org/Ticket/Display.html?id=103765 Tony |
@tonycoz - Status changed from 'open' to 'resolved' |
Migrated from rt.perl.org#123879 (status was 'resolved')
Searchable as RT123879$
The text was updated successfully, but these errors were encountered: