t/op/sprintf.t and t/pmc/sub.t fail during 'make coretest'

[jkeenan]

mikehh reported two test failures (r47147 - Ubuntu 10.04 amd64 (g++)) when running make corevm && make coretest. I got these errors as well on linux/i386.

In discussion on #parrot, bacek suggested that these failures may be side effects of recent migration of ops into dynops. Tests which would previously have passed during make coretest might no longer do so because the functionality they test now depends on things like NQP and PCT, which have not been built at the point make coretest runs. He suggested that some tests might now have to be moved to other files which are not run as part of make coretest.

Could the people who are expert in these matters take a look?

Thank you very much.

kid51

[jkeenan]

Please review the patch attached, immediate_code_as_const.diff, which should correct the failure during 'make coretest' of t/pmc/sub.t.

I have moved the failing test to what seems like a plausible alternative: t/dynoplibs/trans.t. However, I haven't written too many tests in PIR, so we need code review before applying the patch.

Thank you very much.

kid51

[jkeenan]

patch for problem in t/pmc/sub.t only

[plobsing]

Replying to jkeenan:

I have moved the failing test to what seems like a plausible alternative: t/dynoplibs/trans.t. However, I haven't written too many tests in PIR, so we need code review before applying the patch.

The PIR appears to be good, but the intent of the test is to test the behaviour of :immediate :anon subs. The use of the atan dynop is merely to create a known value (namely Pi). I've changed the offending sub.t test in r47167 to use Phi in stead.

[plobsing]

Replying to jkeenan:

mikehh reported two test failures (r47147 - Ubuntu 10.04 amd64 (g++)) when running make corevm && make coretest. I got these errors as well on linux/i386.

The last test t/op/sprintf.t uses the sysinfo op (moved to dynop) to skip certain tests on specific platforms.

To resolve this, we can:

• move the OS-dependant tests out of core ops (eg: t/dynoplibs/sprintf.t)
• remove these tests
• eliminate system dependant behaviour in sprintf

[jkeenan]

Replying to plobsing:

I've changed the offending sub.t test in r47167 to use Phi in stead.

Thanks. I can confirm that that works, leaving only the sprintf problem.

[mikehh]

These tests have now been sorted out and do not fail in make corevm/make coretest

If there is no objection I will close this ticket in the next 24 hours.

[jkeenan]

Confirmed on linux/i386. Confirmed on darwin/ppc.

[mikehh]

Closing this ticket

All tests PASS (pre/post-config, make corevm/make coretest, smoke (#34306), fulltest) at r47544 - Ubuntu 10.04 i386 (g++ with --optimize)

and all other platforms/options that I am aware of