While this isn't something we have tests for in TestBackwardsCompatibility (that only tests every previous version against the current version), we do have tests in TestVersion for parsing versions that do not have constants (see testForwardsCompatibility). Version constants are only shortcuts to Version objects with known values, not what are passed around.
On Tue, Jan 12, 2016 at 8:54 PM, Anshum Gupta <ans...@anshumgupta.net> wrote: > I am not the best person to comment on weather that would block the > release but let me know if you we should. Also, I agree with yonik, we > shouldn't have this limitation as it stops us from pushing out bug fix > releases for past releases. > > On Wed, Jan 13, 2016 at 2:31 AM, Yonik Seeley <ysee...@gmail.com> wrote: > >> On Tue, Jan 12, 2016 at 2:23 PM, Adrien Grand <jpou...@gmail.com> wrote: >> > While working on the 5.4.1 release I just realized that we would not >> prevent >> > users from upgrading from 5.3.2 to 5.4.0 while it is something that we >> don't >> > test in TestBackwardsCompatibility (since 5.4.0 is already released) and >> > that Lucene50SegmentInfoFormat would deserialize a version constant that >> > does not exist among the Version constants. I don't know if this should >> > block this release but I think it's at least annoying... I'm afraid that >> > some of our code might rely on Lucene50SegmentInfoFormat to deserialize >> a >> > version that actually exists? >> >> Seems like this versioning limitation should be fixed - we should >> always be free to create bugfix releases for past releases. >> >> -Yonik >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org >> For additional commands, e-mail: dev-h...@lucene.apache.org >> >> > > > -- > Anshum Gupta >
