gcj on Windows ToDo List

Fri Nov 17 22:08:00 GMT 2006

Mohan Embar wrote:
> Hi Marco,
>> Here is the MinGW ToDo list you sent me off list:
>>>> Between the two of you, could both of you figure out what my initial work
>>> queue will be and prioritize it? I already know about Andrew's park() and
>>> unpark() stuff as well as the three things that Marco mentioned:
>>>>>> http://gcc.gnu.org/ml/java/2006-09/msg00049.html
>> I know of:
>> 1. Serialisation is broken
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28531

I'm working on that...
>> 2. Runtime.exec(String[] cmdarray, String[] envp) doesn't use envp
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29151
>> 3. shared is broken
>> maybe there's a bugreport from someone else
>> 4. Exceptions do no use DWARF
>> maybe there's a bugreport from someone else
>> 5. cygwin compilation is broken
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29476
>> threads don't work either. I think we should concentrate on mingw.
>> 6. JNI broken with certain libs
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29477
>> I can't provide a minimal example, the lib is closed source.
>> 7. using @sourcelist-file leads to wrong flag creation
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29772
>> Andrew, you told me I should tell you ;-)

That's not mingw specific. That's happens on all platforms...
>> 8. tls was changed and broke compilation of windows
>> http://gcc.gnu.org/ml/gcc-patches/2006-10/msg01019.html
>> --disable-tls fixes it but still, this should be fixed
>> Does this still look correct?

Yes, very nice summary. Thanks...
The most important right now are 4 and 6. 6 is - IMHO - a major 
regression. 4 should really slowly be attended to...
> And am I doing this against 4.2 or 4.3?

It's 4.2 and thus also 4.3 because no one changed anything there in the 
mean time (unless I missed something).
> (I tried compiling 4.3 last night but it failed with
> an unsatisfied dependency on MPFR on Fedore Core 5. I just tried
> this for 10 min., though.)

Sorry, I haven't tried the 4.3 yet... I'm actually currently fighting 
with the cross-compilation (host=linux, target=mingw) of the 4.2 branch.
Marco