I think sometimes fingerprints map can consume large amounts of memory, but for not long time. SoftReference allows allocated memory will be released if another proccess will query more memory.
On Thursday, November 8, 2012 5:33:53 PM UTC+2, qazwart wrote: > > So fingerprinting is memory intensive. Storing thousands of fingerprints > for every build isn't going to work. > > -- > David Weintraub > da...@weintraub.name <javascript:> > > On Nov 8, 2012, at 9:58 AM, AdvanTiSS <adva...@gmail.com <javascript:>> > wrote: > > After digging code deeper, i've found that fingerprints are instances of > SoftReference<Fingerprint> class > From javadoc: > ** Soft reference objects, which are cleared at the discretion of the > garbage* > * * collector in response to memory demand. Soft references are most > often used* > * * to implement memory-sensitive caches.* > > > On Thursday, November 8, 2012 2:35:45 PM UTC+2, qazwart wrote: >> >> I understand how the fingerprinting process works and I use it to mark >> our jars, wars, ears, and zips. >> >> However, I was thinking of fingerprinting every file in a zip for every >> build. This way, we can trace back each file when the archive is unzipped >> and its contents scattered. That could mean finger printing hundreds or >> thousands of files in each build. >> >> I suspect that may add a few minutes to the build, but how will tracking >> all those finger prints affect Jenkins performance? >> >> -- >> David Weintraub >> da...@weintraub.name >> >> On Nov 8, 2012, at 3:09 AM, AdvanTiSS <adva...@gmail.com> wrote: >> >> Fingerprinting process based on md5 checksum calculation >> using java.security.DigestInputStream on each file targeted for >> fingerprinting. >> You can read some information about md5 algorithm performance here - [Secure >> hash functions in >> Java<http://www.javamex.com/tutorials/cryptography/hash_functions_algorithms.shtml> >> ]. >> >> On Wednesday, November 7, 2012 3:44:11 PM UTC+2, qazwart wrote: >>> >>> How resource intensive is fingerprinting? What if I fingerprint all >>> files that I build? >>> >>> We deploy a lot of zipped archives instead of jars and wars in our JBoss >>> instant. This way, we can generate various client configurations. However, >>> it also means that the build assets can get moved around quite a bit, and >>> I'd like someway of determining what build that file was associated with. >>> Right now, I'm just fingerprinting the zipped archive, but it may be better >>> if I fingerprinted all the files inside the archive before it is zipped. >>> >>> I can't imagine fingerprinting taking up a lot of resources, on a per >>> file basis, but if I am fingerprinting hundreds of files per build, I can >>> imagine it being a problem. >>> >>> What is your policy on fingerprinting files? >> >>
