[rfc] [icedtea-web] make PAthsAndFilles to follow seted paths IV - bubble to CacheLruWrapper

[Jiri Vanek]

Seems like I forgot an attachment...

On 03/31/2015 03:51 PM, Jiri Vanek wrote:
> There are/were/reappeared two issues in LRUwrapper and CacheUtils
>
> First the file - however crated was as value,not as pointer. So whatever happened to original, had not bubbled to lruwrapper/utils.
> Second - the methods were using getters on field, whcch could change during the method invocation.
>
> This pathc should fix both of them.
>
> The main refactoring is
> -    public CacheLRUWrapper(final File recentlyUsed, final File cacheDir) {
> -        recentlyUsedPropertiesFile = new PropertiesFile(recentlyUsed);
> +    public CacheLRUWrapper(final InfrastructureFileDescriptor recentlyUsed, final InfrastructureFileDescriptor cacheDir) {
> +        recentlyUsedPropertiesFile = recentlyUsed;
>
> Now the cache lru wrapper is handling the "pointers" to sources. So any call to its "getFiles" get real value fromm PathsAndFiles
> When any method is using its getFile it alwasy save current copy, co during runtime, it will not be affected by possible change.
>
> The exception is
>
> -    public PropertiesFile getRecentlyUsedPropertiesFile() {
> -        return recentlyUsedPropertiesFile;
> +    PropertiesFile getRecentlyUsedPropertiesFile() {
> +        if (cachedPropertiesFile == null) {
> +            //no properties file yet, create it
> +            cachedPropertiesFile = new PropertiesFile(recentlyUsedPropertiesFile.getFile());
> +            return cachedPropertiesFile;
> +        }
> +        if (recentlyUsedPropertiesFile.getFile().equals(cachedPropertiesFile.getStoreFile())){
> +            //he underlying InfrastructureFileDescriptor is still poinitng to the same file, use current proeprties file
> +            return cachedPropertiesFile;
> +        } else {
> +            //the InfrastructureFileDescriptor was set to different location, move to it
> +            //is there something needed to close previous instance?
> +            cachedPropertiesFile.store();
> +            cachedPropertiesFile.unlock();
> +            //above were just shotted here
> +            cachedPropertiesFile = new PropertiesFile(recentlyUsedPropertiesFile.getFile());
> +            return cachedPropertiesFile;
> +        }
> +
>
>
> When something change value path to properties file *and* call getRecentlyUsedPropertiesFile then the file will store and unlock. (note - I dont know for sure if those two are needed, but afaik yes). So actions upo  previous Prop.File  may misbehave.
> However .. The setter to the RECENTLY_USED_FILE and KEY_USER_CACHE_DIR is used *only* in tests (and should never be used in ITW itself)
> Afaik - whole those will work if you run javaws (keep running) change proeprties, run second javaws. I noticed no issues. Everything was correctly cached  twice
>
>
> There is one aditional refactoring in this patch - sorry for that - InfrastructureFileDescriptor moved from inner class to outer class. With one change -
> +    //simple constructor to allow testing instances base don overrides
> +    protected InfrastructureFileDescriptor() {
> +        this("nope", "nope", "nope", "nope");
> +    }
>
>
> At the end, I consider this 4 pathces + lrucaches singleton + PAthsAndFiles as reallly good thing which happend to itw.
>
>
> Jie - Is this enough for your "[rfc][icedtea-web] Use temporary cache in *" or is there something more to be done?
>
>
> J.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: madeLRUcahceWrapperToUseInfrastructureFileDescriptorDirectly.patch
Type: text/x-patch
Size: 28395 bytes
Desc: not available
URL: <http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20150401/0046cff7/madeLRUcahceWrapperToUseInfrastructureFileDescriptorDirectly-0001.patch>