TestHarness should be TESTPROD instead of PROD

Bug #1842086 reported by Dirk Zimoch
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
EPICS Base
Status tracked in 7.0
7.0
Low
Andrew Johnson

Bug Description

Several "TestHarness" vxWorks and RTEMS products are built as PROD_vxWorks and PROD_RTEMS instead of TESTPROD_vxWorks and TESTPROD_RTEMS even though they depend on TESTLIB code. This seems a bit inconsistent. Changing them to TESTPROD would allow to build EPICS base without tests, which could speed up the development cycle in an early phase when building the tests is not yet needed.

Attached is a patch for EPICS 7 base. The same could be done for the submodules.

Revision history for this message
Dirk Zimoch (dirk.zimoch) wrote :
Revision history for this message
Andrew Johnson (anj) wrote :

When we first started running the libCom tests on VxWorks I found it useful for the test harness munch file to be installed into the bin/<arch> directory instead of buried somewhere inside the source tree. This is no longer as advantageous since the tests now depend on various other files (mostly .dbd and .db) which are also spread around the source tree (although maybe not quite as much), so I have to cd around the tree anyway to be able to run them.

We don't currently have a way to build Base without also building the tests, but that's actually a completely orthogonal issue. If we wanted that I would probably implement it by adding a TEST_DIRS variable (analogous to DIRS) and only descend into those directories when actually building tests. We only have one TESTPROD being built in a non-test directory at the moment:

Entering 'pva2pva'
p2pApp/Makefile:TESTPROD_HOST += testmon

Changed in epics-base:
status: New → In Progress
importance: Undecided → Low
Revision history for this message
Dirk Zimoch (dirk.zimoch) wrote :

> We don't currently have a way to build Base without also building the tests,
Well, I simply tried to call 'make TESTS= TESTLIB= TESTPROD=' to speed up compilation while implementing my changes to iocLogClient.

Revision history for this message
Dirk Zimoch (dirk.zimoch) wrote :

Same patch for pvData submodule

Revision history for this message
Dirk Zimoch (dirk.zimoch) wrote :

Same patch for pvDatabase submodule

Revision history for this message
Dirk Zimoch (dirk.zimoch) wrote :

Same patch for pvAccess submodule

Revision history for this message
Andrew Johnson (anj) wrote :

These patches should have been applied for this release, sorry this fell through the cracks.

Revision history for this message
mdavidsaver (mdavidsaver) wrote :

> tests now depend on various other files

The RTEMS test harness executables remain self-contained, with necessary test files baked into the executable.

Personally, I don't see skipping re-linking test executables as a major slow down. At least when compared with compiling c++ code like libpvAccess.

@Andrew, Have you committed any of these changes? If not, then I would rather defer this issue again.

Revision history for this message
Andrew Johnson (anj) wrote :

I have not committed or tested any of these yet. I'll remove the milestone.

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers