Crash with "assert(number_of_codes <= 256) failed: too many bytecodes"

[Maurizio Cimadamore]

On 28/07/17 22:15, John Rose wrote:
> On Jul 27, 2017, at 8:57 AM, Tobias Hartmann 
> <tobias.hartmann at oracle.com <mailto:tobias.hartmann at oracle.com>> wrote:
>>
>>>
>>> Also, why did I not see the issue when running all tests?
>>
>> Did you run with a debug build? Maybe your build cached some files. 
>> Re-building after changing bytecodes.hpp should take a long time.
>
> A test that fundamental (and cheap) should fail with "java -version" 
> in a release build.
> If that's not the case, there should be this in our code: 
>  guarantee(number_of_codes <= 256, "").
>
Right - I'm too under the impression that this business of critical 
error turned off in product build (while I might understand some of the 
rationale), seems to be counterproductive - or at least it was in this 
case big time.

I was able to run all jdk and langtools jtreg test, and all JDK/hotspot 
valhalla tests w/o a single glitch, which does seem weird.

That said, if the 'norm' for HS is to run tests in fastdebug mode (or do 
you need full debug to enable these?), then I should just have done 
that, of course.

Maurizio