Bug #15810: Windows spec failure - Process.ppid returns the process id of the parent of this process - ??

• Status: Closed
• Assignee: -
• Target version: -
• Backport: 2.4: UNKNOWN, 2.5: UNKNOWN, 2.6: UNKNOWN

[MSP-Greg (Greg L)]

Since being committed at 79671ec on 28-Apr, this spec has consistently failed:

Process.ppid returns the process id of the parent of this process FAILED
Expected "11424\n"
 to equal "13044\n"

C:/projects/ruby/spec/ruby/core/process/ppid_spec.rb:5:in `block (2 levels) in <top (required)>'
C:/projects/ruby/spec/ruby/core/process/ppid_spec.rb:3:in `<top (required)>'

But, testing it locally, both 'serial' & parallel, it passes. The test also passes at ruby/spec, and it passed with ruby-loco (which doesn't use make)...

As to testing locally, if anyone wants to verify (on Windows), from the spec/ruby folder:

../mspec/bin/mspec -rdevkit  -j -T "--disable=gems" core

[usa (Usaku NAKAMURA)]

• Status changed from Open to Closed

in test-spec target of make, RUBY_EXE constant in mspec is RUNRUBY.
it means that the invoked ruby process is the grandchild of the invokee, not the child.
so, this spec is never passed on Windows via make test-spec.

[MSP-Greg (Greg L)]

@usa (Usaku NAKAMURA)

I revised appveyor.yml to run specs from the install folder, and submitted a PR. I think running at least one test suite from it is a good idea, as the purpose is to create an install.

So, I reverted the commit that bypassed the failing spec. Maybe @Eregon (Benoit Daloze) can create a guard that considers both windows and whether RUNRUBY exists...

[Eregon (Benoit Daloze)]

• Related to Bug #15812: Run specs from install folder? added

[Eregon (Benoit Daloze)]

Would it be possible to use execve() in RUNRUBY, such that RUNRUBY and the actual ruby process have the same PID, or is that not available on Windows?