Java 7 Ships With Severe Bug

Lisandro writes "Lucid Imagination just posted an announcement about a severe bug in the recently released Java 7. Apparently some loops are mis-compiled due to errors in the HotSpot compiler optimizations, which causes programs to fail. This bug affects several Apache projects directly — Apache Lucene Core and Apache Solr have already raised a warning, noting that the bug might be present in Java 6 as well."

Comments

They released this anyway

[Tridus]

Relevant part:

These problems were detected only 5 days before the official Java 7 release,
so Oracle had no time to fix those bugs, affecting also many more
applications. In response to our questions, they proposed to include the
fixes into service release u2 (eventually into service release u1, see [6]).
This means you cannot use Apache Lucene/Solr with Java 7 releases before
Update 2! If you do, please don't open bug reports, it is not the
committers' fault! At least disable loop optimizations using the
-XX:-UseLoopPredicate JVM option to not risk index corruptions.

If this was known before the release and it's as severe as it's being made out to be, why the hell didn't they postpone the release? It's not like the world is dependent on Java 7 being released on time.

This isn't a little issue, either. It's extremely irresponsible for Oracle to put this kind of release out knowing of a bug this severe without any kind of warning on it.

Re:Sounds just about right for Oracle.

[dgatwood]

And if it's such an obvious bug, one would think it would have turned up in the last 29 version 6 releases?

No, honestly. This wasn't caught before because nobody used those flags. Oracle decided that these flags should be turned on by default. Therefore, the onus was on Oracle to thoroughly and broadly test these flags before promoting them to be used by default.

I guarantee you'll find some hairy bugs if you enable lots of random, rarely enabled flags in just about any compiler. The difference between a good compiler and a bad compiler is that a good compiler tests flags thoroughly before either enabling any the flags by default or rolling them into a commonly used option. In effect, what Oracle did was to take an obscure, poorly tested code path and promote it into the hot path through their code. This is something that any first-year CS student should know is risky.

The best part of this is that (assuming other Slashdot comments are correct) this occurs in commonly used third-party libraries, and was disclosed to Oracle several days before the release shipped. Where I work, that's what is known as a P1 block-ship bug, and people will be called in to work on it day and night until the problem is resolved, and if necessary, features will get temporarily pulled (e.g. turning that optimization back off by default).

For shame, Oracle.