On 11/11/2013 02:20 AM, Magnus Ihse Bursie wrote:
On 2013-11-08 16:54, Jonathan Gibbons wrote:
On 11/08/2013 12:22 AM, Magnus Ihse Bursie wrote:
If we can achieve both the goal of moving the webrev script away
from being tied to a specific jdk repo, and still let everyone keep
their scripts/command lines/etc intact, I can't see the harm in
that. It's no big deal forcing people to clone yet another repo to
use webrev, but on the other hand, it's no big deal to make this
change completely transparant as well -- and then I'd prefer the
latter.
I have no problem with another code-tools repo, but it does raise the
question of how folk should get tools provided in the such repos. I
presume you would not suggest that if someone wants to use jtreg,
some script should automagically download and build jtreg. What about
jcheck? If someone wants to use jcheck locally, should the OpenJDK
forest provide a script to download and install jcheck somewhere?
Actually, why could you not have a script to download and build jtreg?
I get the feeling you want to provide a reductio ad absurdum, but for
me, that is not absurd. In fact, that is the direction the new build
system has been going all the time. Providing a configure step, which
finds what is available, and suggests command lines for getting what
is missing, is the first step. Further along, we still want to provide
the "build deps" solution -- where the tools needed to build the JDK
can be automatically downloaded and installed. Sure, this is not
trivial (that's why it's not implemented yet), but that does not mean
it's not a worthy goal.
If we had an existing build-deps infrastructure, it would be a
no-brainer to add webrev (or jtreg) to it. But we do not have it, yet.
So, does that mean we should not try to download anything, or should
we try some tools without have a solution for everything. I'd vote for
the latter.
/Magnus
Making it easier to get a copy of jtreg and other tools, is a worthy
goal, and may even involve building those tools when necessary. But, I
don't think it is sensible to clone "additional" repos into an OpenJDK
forest. Earlier, this thread started because you didn't want to break
existing scripts, but you are as likely to break other scripts if you
start cloning repos from other projects anywhere under the root of an
OpenJDK forest.
Furthermore, whilst I accept that some folk prefer to build from all
source on up, others may prefer to access and use prebuilt binaries when
available.
-- Jon
