Source that compiles with many different compilers is generally a
good thing, assuming the code hasn't been ifdef'd to death.
But successful compilation is just part of the picture.
Anyone delivering built openjdk bits,
needs to have tested those bits and that they understand the
quality issues that a built jdk has on the C/C++ compiler used.
It would be impossible for any one group to guarantee that every
change is buildable with every system & compiler combination,
much less guarantee it runs perfectly in all cases.
But I think that's understood.
I see no reason to block changes that allow our openjdk source to be built
with any reasonable compiler, and even well documented workarounds
could even be justified. But we may want some reasonable limitations
so that the code stays maintainable.
-kto
Andrew Haley wrote:
David Holmes - Sun Microsystems wrote:
Andrew,
Andrew Haley said the following on 06/20/08 20:55:
The changes we had to make to build OpenJDK with gcc 4.3 were to fix
nonstandard C++ code and to turn off -Werror because gcc 4.3 is much
more fulsome in its warnings. Rather than insist on using an older
compiler, we would probably be better off fixing the code that
generates the warnings.
It isn't compilation that is the issue (we should certainly fix the
OpenJDK code to compile cleanly on a standards-compliant compiler) but
the actual runtime behaviour of the compiled code. WE don't "recommend"
the old compiler because it allows old sloppy code to get through, but
because we've already uncovered the bugs that affect us at runtime,
through literally years of use and testing.
Sure. I was counselling against the (oft-seen) practice of using an old
compiler because the code doesn't run on the new one.
Andrew.
