Removal of the old build system, partial review
Magnus Ihse Bursie
magnus.ihse.bursie at oracle.com
Thu Oct 31 14:17:07 UTC 2013
On 2013-10-31 13:28, Alan Bateman wrote:
> On 31/10/2013 07:36, Magnus Ihse Bursie wrote:
>>
>> I'm not sure there is a definite plan for this. Personally, I'd
>> prefer to have this fixed in JDK8, but I'm also aware that the timing
>> is not ideal. Unfortunately, some groups were still using the old
>> build system until very recently, so we have been unable to do this
>> before.
> The timing with respect to the jdk8 schedule is indeed awkward. So if
> you doesn't go ahead then would it make sense to at least disable it
> so as to flush out any remaining slackers? Also is there really risk
> to your proposed step 1 that simple deletes the old make files that
> are no longer used?
The risk to deleting the old make files that are not used, is that I
have not managed to track all uses. For instance, there are tests in
jdk/test/closed that reference files in jdk/make as part of their
testing. (I don't think that's the way it should work, but hey, that's
the way it is.)
Of course we could "disable" the old build slighly more than now, by not
accepting the NEWBUILD argument anymore. This will make it harder (but
not impossible) to build using the old build system.
Still, to a not initiated developer, the old make system will still seem
present and functional. Someone is bound to try to build with it.
Someone is bound to try to update code when making corresponding changes
in the new build system. Etc.
So there are several steps that can be taken:
1) just "disable" to normal way of being able to build using the old
build system
2) just remove parts of the old build that is deemed safe to remove
3) fully remove all unused parts of the old build, and move the new
build into its place
We can of course decide that we should take step 1, or maybe even 2, but
not 3. But I'm not sure what the benefits are. Between today and 1,
there's just those stubbornly enought to specify NEWBUILD=false. We have
no such users internally at Oracle, and I doubt anyone outside Oracle
does this. For 2, we still have the risk of inadvertently deleting
files, but we will end up with a half-baked mess of the old build system
being neither completely there nor completely gone.
So my suggestion is that we either stay on 0, or go all the way to 3.
/Magnus
More information about the build-dev
mailing list