webp is leaking memory

[Antoine Martin]

See ticket:457#comment:30 - this is a client side leak only (I think), which affects all platforms.

Probably keeping references to some buffers somewhere (create_string_buffer?)

Fix it so we can actually use webp (and fix #487), and use it for more things. (ie: #419)

r5167 adds a big warning for anyone trying to use webp, will need to be removed if/when we figure this out

[Antoine Martin]

re-scheduling all webp stuff

[Antoine Martin]

• leak fixed in r5945
• remove leak warnings and re-add webp as a valid lossless encoding to switch to in r5951
• refuse to use webp unless the client tell us it is not going to leak in r5955

All will need to be backported.

webp support via PIL was added in:

• r5937 for encoding
• r5936 for decoding (r5953 ensures we don't use Pillow for decoding webp for now - we can lift this restriction with a version check once upstream fixes the bug)

Pillow also leaks like we used to! (see r5952)

Bug reported upstream: ​Pillow issue 586

[Antoine Martin]

Backports for v0.12.x in r5956. Closing.

[Antoine Martin]

Note: the new webp cython encoder in 0.13 was also leaking memory, this is fixed in r6505.

[migration script]

this ticket has been moved to: ​https://github.com/Xpra-org/xpra/issues/491