[8u] RFR 8216401: Allow "file:" URLs in Class-Path of local JARs

[Andrew John Hughes]

On 13/08/2019 20:34, Zhengyu Gu wrote:
> Hi Clive,
> 
> On 8/13/19 2:59 PM, Verghese, Clive wrote:
>> Hi Zhengyu,
>>
>> Instead of importing the writeToDisk function to the test, We could
>> update the ClassFileInstaller to be similar to the one in tip. [1]
>>
>> http://cr.openjdk.java.net/~phh/8216401/webrev.8u0.00/
> 
> I think the principal of 8u backport, is to minimize the scope to
> absolutely needed.

Not necessarily. It varies on a case-by-case basis and it's usually
worth doing a little repository archaeology to determine why differences
exist.

It's less about the most expedient way to do a single backport as to the
long term maintainability of the source code base. If there's an earlier
change that will make the original patch apply more easily, the
advantages of doing so, and the benefits of that to later backports,
needs to be weighed against the risks of backporting that dependency.

The refactoring we backported with the printing implementation last time
is a good example. The risk of renaming those internal classes is
minimal - especially as the new names are already in use in later JDKs -
and future backports in that area are now more likely to apply cleanly.

 I imported some of ClassFileInstaller code to the
> test, because I have not seen they are used by others. If that's the
> case, I think we should arrange partial backport for test library
> changes. I don't like the idea to sneak in test infrastructure changes
> via unrelated backports.

I agree, though it should be done under the original changes as much as
possible. See my comments on JDK-8229501.

> 
> Thanks,
> 
> -Zhengyu
> 


-- 
Andrew :)

Senior Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: ed25519/0xCFDA0F9B35964222 (hkp://keys.gnupg.net)
Fingerprint = 5132 579D D154 0ED2 3E04  C5A0 CFDA 0F9B 3596 4222
https://keybase.io/gnu_andrew