Stream: git-cranelift

Topic: cranelift / PR #1332 Try to reduce CI times with a Rust `...

GitHub (Jan 10 2020 at 21:51):

alexcrichton opened PR #1332 from wabt-to-wat to master:

This commit moves the cranelift tests and tools from the wabt crate on
crates.io (which compiles the wabt C++ codebase) to the wat crate on
crates.io which is a Rust parser for the *.wat format. This was
motivated by me noticing that release builds on Windows are ~5 minutes
longer than Linux builds, and local timing graphs showed that wabt-sys
was by far the longest build step in the build process.

This commit changes the clif-util binary where the --enable-simd
flag is no longer respected with the text format as input, since the
wat crate has no feature gating. This was already sort of not
respected, though, since --enable-simd wasn't consulted for binary
inputs which clif-util supports as well. If this isn't ok though then
it should be ok to close this PR!

GitHub (Jan 10 2020 at 21:55):

alexcrichton updated PR #1332 from wabt-to-wat to master:

This commit moves the cranelift tests and tools from the wabt crate on
crates.io (which compiles the wabt C++ codebase) to the wat crate on
crates.io which is a Rust parser for the *.wat format. This was
motivated by me noticing that release builds on Windows are ~5 minutes
longer than Linux builds, and local timing graphs showed that wabt-sys
was by far the longest build step in the build process.

This commit changes the clif-util binary where the --enable-simd
flag is no longer respected with the text format as input, since the
wat crate has no feature gating. This was already sort of not
respected, though, since --enable-simd wasn't consulted for binary
inputs which clif-util supports as well. If this isn't ok though then
it should be ok to close this PR!

GitHub (Jan 10 2020 at 21:55):

alexcrichton created PR Review Comment:

It's worth noting here that wat itself has the fallbacks already where if the input file is a wasm binary it won't attempt to parse it as text, it'll just return the raw binary that was read. When I removed the call to read_to_end here it made that a dead function, so I inlined the logic for handling -

GitHub (Jan 10 2020 at 21:55):

alexcrichton submitted PR Review.

GitHub (Jan 10 2020 at 21:59):

abrown submitted PR Review.

GitHub (Jan 10 2020 at 21:59):

abrown created PR Review Comment:

I think then that we should remove the add_enable_simd_flag, add_enable_multi_value, add_enable_reference_types_flag functions.

GitHub (Jan 10 2020 at 22:00):

alexcrichton updated PR #1332 from wabt-to-wat to master:

This commit moves the cranelift tests and tools from the wabt crate on
crates.io (which compiles the wabt C++ codebase) to the wat crate on
crates.io which is a Rust parser for the *.wat format. This was
motivated by me noticing that release builds on Windows are ~5 minutes
longer than Linux builds, and local timing graphs showed that wabt-sys
was by far the longest build step in the build process.

This commit changes the clif-util binary where the --enable-simd
flag is no longer respected with the text format as input, since the
wat crate has no feature gating. This was already sort of not
respected, though, since --enable-simd wasn't consulted for binary
inputs which clif-util supports as well. If this isn't ok though then
it should be ok to close this PR!

GitHub (Jan 10 2020 at 22:00):

alexcrichton submitted PR Review.

GitHub (Jan 10 2020 at 22:00):

alexcrichton created PR Review Comment:

Done!

GitHub (Jan 10 2020 at 22:31):

abrown submitted PR Review.

GitHub (Jan 10 2020 at 22:32):

abrown merged PR #1332.

Last updated: Dec 23 2024 at 13:07 UTC