I'm for making it obvious and keeping clean ups localized, meaning add try-with-resources blocks to all close() methods that need them.
Gary On Thu, Jun 28, 2018 at 3:54 AM Stefan Bodewig <bode...@apache.org> wrote: > Hi all > > https://issues.apache.org/jira/browse/COMPRESS-457 raises an issue that > I vaguely recall we've talked about in the past but I may be wrong. > > Almost all our OutputStream close methods go along the lines of > > public void close() throws IOException { > finishFormatSpecificStuff(); > closeUnderlyingStreamsOrChannelsAndOtherResources(); > } > > as some formats need to write trailers in order to create valid > output. If for some reason finishFormatSpecificStuff() stuff fails the > underlying stream will not be closed leading to a resource leak - unless > you keep track of the underlying stream and close it yourself > (try-with-resources probably). > > Do we want to do anything about this and if so what? > > We could modify all close implementations to perform resource cleanup in > a finally block. Likewise we could add new (let's say unsafeClose()) > methods that only perform resource cleanup. Or we could properly > document the behavior and ensure our examples contain resource cleanup > code. > > I'm leaning toward "documenting" but want to ensure there are no leaks > the user cannot prevent. For example DeflateCompressorOutputStream > "end()"s the Deflater instance in a finally block while zip's > StreamCompressor holds a reference to a Deflater that may never get > closed. > > Stefan > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org > For additional commands, e-mail: dev-h...@commons.apache.org > >
