Fix Windows 7 (Debug) Build on Jenkins

[Peter Parker]

This build was introduced recently and has been failing ever since.

There are three problematic tests:

• AlgorithmsTest.FindPeaksTest
• AlgorithmsTest.StripVanadiumPeaks2Test
• CrystalTest.MaskPeaksWorkspaceTest

Jenkins is low on details but a run of the VS debugger seems to suggest that these are just related to iterators / indexes being out of bounds. Hopefully, relatively easy to fix.

[Peter Parker]

Refs #8028 - Avoid dereferencing an end iterator.

Changeset: 0a83fb72eb69b5cfd8f5680f81f03178a2b381fb

[Peter Parker]

Refs #8028 - Ensure map is being accessed at an existing key.

These changes pass the unit tests, but it would be good if someone who knows more about this code could take a look to double check it works as intended.

Changeset: 0ed1dec9c6fb9341e5d7933a8b4db3a8c5f4f415

[Peter Parker]

Refs #8028 - Pin down debug error with unit test.

Changeset: 928873f16b7b506fd904c61ee6881ea42d658a39

[Peter Parker]

Resolving as fixed.

Vickie confirmed that changeset 0ed1dec9c6fb9341e5d7933a8b4db3a8c5f4f415 was reasonable.

The three unit tests pass on my machine, hopefully that will fix the build on Jenkins.

Tester: make sure the ​Windows Debug build is passing.

[Vickie Lynch]

Refs #8028 fix system test broken by this ticket

Changeset: b6f3c2b61a95856955b6aed7efe247ce9b970e06

[Peter Parker]

I'm reopening this ticket, since PointByPointVCorrectionTest ​failed this morning.

It appears to be a multi threading issue. Sometimes the test fails and sometimes it passes.

@Vickie, I can't seem to see the changeset in ​comment 5. Was that part of this ticket, or added by mistake?

[Peter Parker]

Refs #8028 - Fix sporadically-failing unit test.

• Replaced buggy use of std::bind2nd within PARRELLEL macros.
• Removed unnecessary call to a workspace's dataX() method.

Changeset: 8880bcc23fac78bc1768eb5ac5c763be5d071b1a

[Peter Parker]

Refs #8028 - Skip python unit test on Windows Debug only.

Since we cannot easily support sqlite3 on Windows in Debug, just skip this single test if we can't import sqlite3.

Also removed redundant import check at start of file, since it gets checked in the "test_imports" method later.

Changeset: dc00612ef1b82401ca24ce65adc0cc187ff791ec

[Peter Parker]

Refs #8028 - Manually skip test rather than use skipIf decorator.

"skipIf" was only introduced in Python version 2.7, so we have to skip the test manually since we have an older version of Python on our RHEL 6 build.

Changeset: d091b6f48189145f458e00c2cfc39bdec59ad4c5

[Peter Parker]

Refs #8028 - Remove files I added accidentally.

A "git add ." caught some files I had in my working directory.

Changeset: 745e87f2223a5fc284ade3976913f1863a37b32e

[Peter Parker]

The build appears to be fully fixed now and the system tests were fixed by Vickie. Closing.

Tester:

• Look through my changes and make sure they look appropriate.
• Make sure to merge Vickie's system test changes if you end up passing the ticket.

[Russell Taylor]

The change to PointByPointVCorrection.cpp at lines 88-89 means that the X values are no longer being copied into the output workspace. You need to put in something like outputWS->setX( i, inputWS->refX(i) );

It doesn't say much for the testing of this algorithm that this went unnoticed - be sure to put something in the unit test to ensure this can't happen again.

[Peter Parker]

Refs #8028 - Ensure that X values are carried over to alg result.

• Add setX() line as per Russell's suggestion.
• Cover the X data with a unit test.

Changeset: 9ae2141c878b75f2ec3a313504bc5ecaf4dc40a3

[Peter Parker]

Marking as fixed again.

[Russell Taylor]

• PointByPointVCorrection fix checks out
• The Jenkins build is passing
• I ran these tests through valgrind, which reported plenty of errors before the changes and none afterwards (not sure about MaskPeaksWorkspaceTest - that seemed OK anyway and hasn't been touched in this ticket).
• I'm hoping this will solve the occasional failure of the CalibrateRectangularDetector_Test.PG3Calibration system test, which sometimes crashes in GetDetOffsetsMultiPeaks (which calls FindPeaks)

[Russell Taylor]

Merge remote branch 'origin/feature/8028_fix_windows_7_db_build'

Full changeset: ​cfd8495d2578ce28fcd3b3bb48546c1b31a8b987

[Russell Taylor]

Merge remote branch 'origin/feature/8028_fix_windows_7_db_build'

Full changeset: ​59679062bff320645dc462065845ffbc5c2f34b3

[Stuart Campbell]

This ticket has been transferred to github issue ​8873