llvm2lcov: use ignorable_error rather than die for unsupported MC/DC entries format

[belyaevrd]

No description provided.

[henry2cox]

In this particular case: I think it is (moderately) safe to continue after noticing the issue - unless the array size is less than 10
Or maybe, if the list size is less larger and the 10th element (index 9) is not an array.
That is:
last if $mSize < 10 || ref($m->[9]) ne 'ARRAY'
But that might be too confusing/hard to document/hard for the user to understand what is happening.
So it might be better to just skip the MC/DC entry if the format is not what we expected.
User will then get a later error, when there are no MC/DC coverpoints despite the --mcdc command line argument.

[belyaevrd]

In this particular case: I think it is (moderately) safe to continue after noticing the issue - unless the array size is less than 10

I think that if the MC/DC entries format is unexpected, then these entries might have fields that are very different from the expected ones (i.e., having a new field at the beginning). This can lead to broken results.
So, for me, this exact version checking makes sense.

Or maybe, if the list size is less larger and the 10th element (index 9) is not an array.
That is:
last if $mSize < 10 || ref($m->[9]) ne 'ARRAY'
But that might be too confusing/hard to document/hard for the user to understand what is happening.
So it might be better to just skip the MC/DC entry if the format is not what we expected.
User will then get a later error, when there are no MC/DC coverpoints despite the --mcdc command line argument.

Do you want to remove ignorable_error and just leave the loop via last? I agree with you that it might be too confusing for the user. However, if you want to prevent emitting two errors in such situations, I can remove it.

[henry2cox]

Actually - I think your reasoning above is correct...someone might change the data layout in a way we don't expect - so simply skipping the MC/DC entry after emitting a message is probably the best approah.
It also means that any user who hits the message and really wants MC/DC support, will have to either report the issue or fix it themselves. (Hopefully: sending PR after fix.)

Lets leave this PR as is...I'll merge it tomorrow if nobody else comments between now and then.