Fix tools by removing usages of deprecatedWriter. #25220
Conversation
|
I decided to not touch arocc since it would make sense to update it with the rest of arocc (arocc already removed usages of deprecatedWriter). |
AndrewKraevskii
changed the title
| @@ -122,7 +122,9 @@ fn mode(comptime x: comptime_int) comptime_int { | |||
| } | |||
|
|
|||
| pub fn main() !void { | |||
| const stdout = std.fs.File.stdout().deprecatedWriter(); | |||
| var stdout_buffer: [0x100]u8 = undefined; | |||
There was a problem hiding this comment.
If you're going to flush at every write, instead simply give it an empty buffer. Or, put only one flush at the very end.
Note: this conflicts with another merged patch so you can probably just drop this change.
There was a problem hiding this comment.
I did what you suggested at first but when I checked and print() actually will make multiple syscalls if provided with empty buffer.
Additionally im flushing not after all writes but only before blocking writes e.g. printing results of benchmark and printing header for next benchmark in one syscall since they become available for printing at the same time and it preserves old behavior.
But I guess it doesn't matter for this scripts much so might as well drop it to make implementation simpler.
There was a problem hiding this comment.
You're right, it doesn't matter too much, other than that any code that exists, serves as an example for people reading the code to follow, so there's value to always doing best practices, particularly in ziglang/zig. Not a merge blocker though.
There was a problem hiding this comment.
Im not sure which practice of these two you mean to be best? Doing simpler code (no buffering) or doing faster code (with buffering)?
AndrewKraevskii
force-pushed
the
remove_deprecated_writer
branch
from
acd6cdb to
1c66570
Compare
|
@andrewrk I rebased it on master |
I tried to use
zig reducebut it bitrot a bit since it used oldGenericWriter. I went and updated it and some other tools to use new Writer.Also Adler32 API changed so I also updated lib/std/hash/benchmark.zig to handle it properly just so I can test my changes with Writer.