[roottest] use cmake for `io/evolution` tests

[linev]

There are several subdirs, which are handled fully independent now.

Several common scripts like compile.C or run.C moved to directories where they used - numeric_conversion and stl_conversions

Split several reference files - according to independent tests running

Remove no longer necessary .rootrc, rootlogon.C, rootalias.C files

Finally split large makefile test on 171 separate cmake tests

[github-actions[bot]]

Test Results

    18 files      18 suites   3d 9h 13m 21s ⏱️  3 016 tests  3 016 ✅ 0 💤 0 ❌ 52 835 runs  52 835 ✅ 0 💤 0 ❌

Results for commit a6bde992.

:recycle: This comment has been updated with latest results.

[linev]

I see sporadic errors of roottest-root-io-evolution-fixarr2 on Mac platforms. Also on my Mac I see in 10% this test failing. Not sure if problem cased by changing from Makefile to cmake

[pcanal]

I see sporadic errors of roottest-root-io-evolution-fixarr2 on Mac platforms.

What is the error?

[linev]

I see sporadic errors of roottest-root-io-evolution-fixarr2 on Mac platforms.

What is the error?

See result of last CI on mac15:

https://github.com/root-project/root/actions/runs/15714091698/job/44279370611?pr=19012#step:6:8703

It produces stack trace when file is deleted. In rare cases same failure appears on my Mac and once or twice on my Linux.

Old Makefile-based tests may miss such problem while only stdout was analyzed - which is ok at this moment

[linev]

The variable RootExeOptions and the argument ROOTEXE_OPTS now seems to have the exact same semantic. Any reason to keep both?

For me ROOTEXE_OPTS is much clear compared to RootExeOptions. While RootExeOptions affects all tests in current directory - and subdirs. I rather prefer replace use of RootExeOptions - by ROOTEXE_OPTS. There are very few places in ROOT

[pcanal]

For me ROOTEXE_OPTS is much clear compared to RootExeOptions. I rather prefer replace use of RootExeOptions - by ROOTEXE_OPTS.

I totally agree. Can you make a separate PR with just that change? Thanks.

[pcanal]

Old Makefile-based tests may miss such problem while only stdout was analyzed - which is ok at this moment

The old makefile would not (should not) missed any failure that include a stack trace or crash.

[linev]

The old makefile would not (should not) missed any failure that include a stack trace or crash.

I just checked roottest repository - Makefile not activate fixarr target. So test was not running at all.

[pcanal]

I just checked roottest repository - Makefile not activate fixarr target. So test was not running at all.

Fair enough. We should probably open an issue to remind us to look at it later.

[linev]

For me ROOTEXE_OPTS is much clear compared to RootExeOptions. I rather prefer replace use of RootExeOptions - by ROOTEXE_OPTS.

I totally agree. Can you make a separate PR with just that change? Thanks.

@pcanal I prepared such branch: https://github.com/linev/root/tree/roottest_exe_opts

But first this PR must be merged.