On 08/10/2010 12:08 PM, Dr Andrew John Hughes wrote:
On 12:32 Tue 10 Aug , Göran Uddeborg wrote:
> Andrew Haley:
>> On 08/10/2010 11:10 AM, Alexander Kurtakov wrote:
>>>>
>>>> The sensible rule everyone seems to use is that if adding gcj support
>>>> doesn't require much effort, add it. If not, don't.
>>>
>>> As gcj is at 1.5 JVM level there is no point in having gcj support for
>>> packages that require Java 1.6.
>>
>> Yes, obviously. I don't quite understand the point you're making,
>> though: clearly if a package requires 1.6 then adding gcj support
>> requires much effort, so don't add it.
>
> The source code does not require 1.6. According to the documentation
> of the package, 1.2.2 is enough. (I haven't tried anything less than
> 1.5.)
>
> Originally I had 1.2.2 as a build requirement. When doing a build in
> a minimal environment, like mock, that requirement is met by the GCJ
> compiler. Compiling with the GCJ javac caused very many warnings,
> though, as my reviewer pointed out. The 1.6.0 javac didn't give all
> those warnings.
Warnings or errors? gcj uses ecj under the hood, which is more
verbose by default (more equivalent to javac -Xlint:all). The OpenJDK
source code produces ~10k warnings when compiled this way but still
works. In particular, it warns about generics usage and deprecation
on a per-use basis, whereas javac gives a single warning by default.
It even moans about unused import statements.
Indeed. It's interesting that all this pointless bellyaching ecj
does makes people think something has gone wrong.
Andrew.