[rfc] [icedtea-web] enable process-wide caching on windows

[Jiri Vanek]

On 09/06/2018 02:17 PM, Alex Kashchenko wrote:
> On 09/06/2018 11:13 AM, Jiri Vanek wrote:
>> On 09/06/2018 12:30 AM, Alex Kashchenko wrote:
>>> On 09/05/2018 02:42 PM, Jiri Vanek wrote:
>>>> On 09/05/2018 12:01 AM, Alex Kashchenko wrote:
>>>>> Hi,
>>>>>
>>>>> This patch enables caching of downloaded JARs on windows. Caching is only process-wide - 
>>>>> multiple instances of WebStart apps downloading JARs simultaneously may break caching logic. It 
>>>>> still seems to be better than no caching at all:
>>>> Thats deffinitley true.
>>>>>
>>>>> http://cr.openjdk.java.net/~akasko/itw/windows_cache.patch
>>>>>
>>>> The patch looks good, I'm just wondering, can it be done in more readable way?
>>>
>>> It adds as minimal changes as possible.
>>>
>>>
>>>> Eg having WindowsLockedFile extending LockedFile or simialrly?
>>>
>>> It can be done, but may be non-trivial (OS-wide named mutex looks like an ideal solution for 
>>> that, but it cannot be accessed in pure java). I have no plans to work on this myself.
>>
>> No. No more then workround is necessary here, and no more work is deffinitly implied on you yourself.
>>
>> My issue is, that now two completely unrleated things are mixed in the conditions.
>>
>> Myabe to have two bigger sections of
>>
>> isWindwos   then
>> else
>>
>> would be more readable?  What do you think?
>>
>> The WindowsLockedFile would actually do the same, but should spare the likely appearing dupcode.
>> As the LockedFile already is factory-created, then the WindowsLockedFile inner class of 
>> WindowsLocked  would be really small change.
>>
>>
>> How do you feel abbout attached approach? ***!!!not tested!!!***
> 
> I can suggest to test this first, because caching logic there is quite convoluted.
> 
> Minimal patch was tested some months ago before its inclusion with MSI, I think it is better to push 
> it instead.

Hmm:(( I'm really unhappy with it.
It can possibly go to 1.7, but for head I would like more readable approach which is not mixing 
conditions.
The patch I posted do not touch linux code at all, and windows changes were 1:1 derived from your 
patch. But I could read the conditions wrongly.  Do you mind to try review it theoretically and in 
best to test it for windows?

Sorry for making troubles, but you yourself must admit that the conditions are unrelated.
    J.
> 
> 


-- 
Jiri Vanek
Senior QE engineer, OpenJDK QE lead, Mgr.
Red Hat Czech
jvanek at redhat.com    M: +420775390109