Upgrade miniz_oxide and add unit tests of the different flush kinds #498
Conversation
|
Could we get this done towards the end of the year? Maybe it works now if the PR was rebased? |
|
Right, thanks a lot! I think I may have asked @oyvindln that before and apologise this question is redundant or repetitive, but given my ignorance, is a release of |
|
Yeah I'm hoping to get something out before the end of the year, just need to resolve a few small things. |
fintelia
changed the title
There was a problem hiding this comment.
That's fantastic, thanks so much!
Could we squash this PR into one and, most importantly, add a message that describes the change and ideally the motivation?
While I could do it with the GH UI when merging, I don't like to put words into peoples mouths given that their name is all over it 😅.
This PR adds tests of the various flush kinds to ensure that all backends implement them consistently. The cases are carefully constructed so that even quite different backend implementations are likely to produce the expected output. In particular, it counts on the encoder detecting when the input is/isn't compressible and writing either compressed or stored blocks accordingly. This also upgrades miniz_oxide and corrects it to use partial fushes when requested.
3 participants
This PR adds tests of the various flush kinds to ensure that all backends implement them consistently. The cases are carefully constructed so that even quite different backend implementations are likely to produce the expected output. In particular, it counts on the encoder detecting when the input is/isn't compressible and writing either compressed or stored blocks accordingly.
Tests pass for me with the
zlib,zlib-ng,zlib-rsandcloudflare_zlibbackends. Theminiz_oxidebackend requires Frommi/miniz_oxide#179 as otherwise it treats partial flushes as "no flush" (!)