Update the brotli submodule to its latest version. #179
Conversation
The brotli encoder got a new faster compression mode, quality 0, which is about 2x faster than the fastest mode in the previous version, and the quality 1 mode also got considerably faster. Because of this, it might be worth re-running the brotli benchmarks and updating the benchmark results.
|
How can I reproduce these failures? After I built everything, I tried: $ make test |
|
The unit tests currently use the testing framework built into glib; if you don't have the glib development package install they will be disabled.
Sorry about this, I'm working on removing the glib dependency now. |
|
FWIW, I'm updating all the submodules right now, and this error happens without any changes to squash-brotli.cpp; it appears to be due to a change in brotli itself. If you fire up a debugger you can set a breakpoint on the squash_error function (which is called whenever something returns an error code, just so it's a bit easier to find the source), and a quick backtrace will lead you back to squash-brotli.cpp:219. So that call to WriteBrotliData is failing, and it wasn't before. |
|
I have found the cause and have a fix upstream. After that lands, I will update this PR. |
This is supposed to fix the flush-test failure.
|
The travis failures seem to be unreleated to the brotli update, so brotli should be ok to update. |
|
Merged after a quick rebase, thanks! I'll update the benchmark soon. |
The brotli encoder got a new faster compression mode,
quality 0, which is about 2x faster than the fastest
mode in the previous version, and the quality 1 mode
also got considerably faster.
Because of this, it might be worth re-running the brotli
benchmarks and updating the benchmark results.