Critical: code after EOF is ignored

[mzuenni]

reading the spec I would say that EOF should be identical to ASSERT(ISEOF). However, it seems like the actual implementation stops after encountering EOF. The following checker accepts an empty:

EOF
ASSERT(1!=1)

I would say that is a (critical) error. The ASSERT should be executed and therefore the validator should reject everything.

I have also found at least one checktestdata program that assumed that code / asserts are still executed after the EOF...

[eldering]

reading the spec I would say that EOF should be identical to ASSERT(ISEOF). However, it seems like the actual implementation stops after encountering EOF. The following checker accepts an empty:

EOF
ASSERT(1!=1)

I would say that is a (critical) error. The ASSERT should be executed and therefore the validator should reject everything.

I agree.

I think that behaviour may have been put in place from before when we had ASSERT commands. So there was no point in having anything after an EOF command.

I have also found at least one checker that assumed that code /asserts are still executed after the EOf...

You mean a checktestdata program?