[jdk17u-dev] RFR: 7036144: GZIPInputStream readTrailer uses faulty available() test for end-of-stream

[Todd V. Jonker]

On Fri, 4 Jul 2025 10:36:08 GMT, Goetz Lindenmaier <goetz at openjdk.org> wrote:

>> This is a clean backport of [d3f3011](https://github.com/openjdk/jdk/commit/d3f3011d56267360d65841da3550eca79cf1575b), originally authored by @archiecobbs and reviewed by @jaikiran
>> 
>> The corresponding backport for JDK21 is https://github.com/openjdk/jdk21u-dev/pull/1897
>> 
>> Testing:
>> - [x] New regression test fails without the fix, passes with it
>> - [x] linux-x86_64-server-release: headless tiers 1-3
>
> Hi @toddjonker 
> Have you tested anything besides the jtreg tests mentioned above?
> This change modifies behaviour, this might have negative side effects, especially on the many working installations that receive this update.  Oracle tried to backport, but they didn't deliver it.  Usually, they have a good reason for this.  
> Do we need to rush this backport?  I would rather defer it to the next (January) release to get experience with the change in 21. People can go to 21 or soon 25 if they have an issue.

Hi @GoeLin, I've not tested further. My expectation was that the "bake time" sufficient to backport to 21 was sufficient for earlier releases.

> People can go to 21 or soon 25 if they have an issue.

In the environment I support, we cannot assume that JDK updates, _especially_ from older releases (like 11), are straightforward; in fact they can be quite costly.  I'd personally rather prevent the inevitable-at-scale data loss than tell customers they have to update to avoid it.

Anyway, it sounds like you'd prefer we hold off on this for a while, so I'll update our own issue tracking and then pull this back into draft status.

-------------

PR Comment: https://git.openjdk.org/jdk17u-dev/pull/3667#issuecomment-3046618144