You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@harmony.apache.org by "Alexander Frey (JIRA)" <ji...@apache.org> on 2007/07/23 00:21:31 UTC
[jira] Created: (HARMONY-4514) [drlvm][opt] Wrong line numbers in
stack traces
[drlvm][opt] Wrong line numbers in stack traces
-----------------------------------------------
Key: HARMONY-4514
URL: https://issues.apache.org/jira/browse/HARMONY-4514
Project: Harmony
Issue Type: Bug
Components: DRLVM
Environment: WindowsXP, AthlonXP
Reporter: Alexander Frey
While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4514) [drlvm][jit][opt] Wrong line numbers
in stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Mikhail Fursov updated HARMONY-4514:
------------------------------------
Patch Info: [Patch Available]
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: H4514.java, jit2.patch, TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4514) [drlvm][jit][opt] Wrong line numbers
in stack traces
Posted by "Alexander Frey (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alexander Frey updated HARMONY-4514:
------------------------------------
Attachment: TraceTest.java
I noticed this issue while running the test attached to HARMONY-4513.
I tried throwing some exceptions but couldn't produce wrong line numbers that way.
So I've written another junit test (attached) and again the line numbers reported while running with -Xem:opt didn't match with the line numbers of the assert* methods, while running with -Xem:interpreter the line numbers are correct.
interpreter: 10; 19; 28; 37; 46
opt: 7; 19; 25; 34; 43
The interesting part is the second method in the test: interpreter and opt both report the correct line number (the call of fail()).
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4514) [drlvm][jit][opt] Wrong line numbers
in stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Mikhail Fursov updated HARMONY-4514:
------------------------------------
Attachment: jit.patch
jit2.patch - the fix and regression test.
The problem was introduced by recent bcmap&inlineinfo support refactoring.
After this refactoring every inlined inst have bc-offset in it's original method.
In this patch I keep design the same but changed the way global bc-mapping is calculated: for every inst in inlined method bc-map keeps top-level method's offset (as it was before refactoring).
I also cleaned BCMap code and make it store (uint32, uint16) pairs in image instead of (PINT, PINT). It should reduce bcmap size on IA32 platform from 8 and 16 bytes per entry till 6 bytes on IA32 and EM64T platforms.
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: H4514.java, jit.patch, TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4514) [drlvm][jit][opt] Wrong line numbers
in stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Mikhail Fursov updated HARMONY-4514:
------------------------------------
Attachment: H4514.java
H4514: code for regression testsuite based on Alexanders test
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: H4514.java, TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4514) [drlvm][jit][opt] Wrong line numbers
in stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Mikhail Fursov updated HARMONY-4514:
------------------------------------
Attachment: jit2.patch
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: H4514.java, jit2.patch, TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Assigned: (HARMONY-4514) [drlvm][opt] Wrong line numbers in
stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Mikhail Fursov reassigned HARMONY-4514:
---------------------------------------
Assignee: Mikhail Fursov
> [drlvm][opt] Wrong line numbers in stack traces
> -----------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4514) [drlvm][jit][opt] Wrong line numbers
in stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Mikhail Fursov updated HARMONY-4514:
------------------------------------
Estimated Complexity: Advanced
Summary: [drlvm][jit][opt] Wrong line numbers in stack traces (was: [drlvm][opt] Wrong line numbers in stack traces)
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Resolved: (HARMONY-4514) [drlvm][jit][opt] Wrong line
numbers in stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Mikhail Fursov resolved HARMONY-4514.
-------------------------------------
Resolution: Fixed
committed as r559409
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: H4514.java, jit2.patch, TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4514) [drlvm][opt] Wrong line numbers in
stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12514597 ]
Mikhail Fursov commented on HARMONY-4514:
-----------------------------------------
The reasons are:
1) Not all NPE points have valid bc-mapping. BC-mapping is complete in HIR but can be missed in LIR on some NPE points. Need simple test-cases to found such places and fix. The fix will be easy.
2) Running Harmony on AthlonXP enables extra 'i586' pass that lowers SSE2 ops to x87. AFAIK nobody test this mode today, so the reason of NPE can be a bug in 'i586' pass, but not insufficient bc-mapping.
> [drlvm][opt] Wrong line numbers in stack traces
> -----------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Closed: (HARMONY-4514) [drlvm][jit][opt] Wrong line numbers
in stack traces
Posted by "Alexander Frey (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alexander Frey closed HARMONY-4514.
-----------------------------------
Thanks Mikhail.
Verified at r560888.
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: H4514.java, jit2.patch, TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HARMONY-4514) [drlvm][jit][opt] Wrong line
numbers in stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12514844 ]
Mikhail Fursov commented on HARMONY-4514:
-----------------------------------------
Alexander,
thank you for the test - I'm going to fix this issue in a week and add your test into our regression suite.
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HARMONY-4514) [drlvm][jit][opt] Wrong line numbers
in stack traces
Posted by "Mikhail Fursov (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HARMONY-4514?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Mikhail Fursov updated HARMONY-4514:
------------------------------------
Attachment: (was: jit.patch)
> [drlvm][jit][opt] Wrong line numbers in stack traces
> ----------------------------------------------------
>
> Key: HARMONY-4514
> URL: https://issues.apache.org/jira/browse/HARMONY-4514
> Project: Harmony
> Issue Type: Bug
> Components: DRLVM
> Environment: WindowsXP, AthlonXP
> Reporter: Alexander Frey
> Assignee: Mikhail Fursov
> Attachments: H4514.java, TraceTest.java
>
>
> While running some JUnit tests with the latest snapshot (r557495) and Eclipse 3.3 I noticed wrong line numbers in the stack traces of the failed tests. It seems always the first line of the test method is reported instead of the failing line.
> The issue occures w/o parameters and with -Xem:opt. I can't test it with -Xem:jet because java returns immediately w/o running or reporting any error (issue with Harmony on WinXP+AthlonXP?).
> Running the same tests with -Xem:interpreter or with the M2 build the correct lines are reported.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.