Closed piyush-kurur closed 5 years ago
@Mistuke I am not sure why but I have tried running benchmarks and it fails on appveryor. (Note, it is not a compilation failure). Do you have any pointers for fixing this ?
Do your benchmarks test Unicode characters?
Yeah okey I will supress that
You don't have to, change the console's codepage to UTF8. Windows's default codepage isn't a Unicode one.
Add
chcp 65001
somewhere in your AppVeyor script before running the tests.
That should work, if it doesn't then you may need to tell GHC about it too
hSetEncoding stdout utf8
though this shouldn't be needed as GHC defaults to the codepage of the console.
looks like you do need to set the encoding type in the tests as well. When the new I/O manager is in this should work out of the box, but for now you have to manually change the encoding.
@Mistuke thanks for the prompt help. Now the build is fixed.
@piyush-kurur no problem :) btw I saw this in the logs
Raaz.Core.ByteSource
/dev/null
should return Exhausted even for a read request of 0
# PENDING: Non-posix system needs an equivalent of /dev/null
Windows's NULL device is called NUL
.
For GHC's prior to 8.6 you can access it using just NUL
as the file name, so e.g. /NUL
, for GHC's after and including 8.6 you need \\.\NUL
(escaped as \\\\.\\NUL
of course) as the filename. You'll need to ifdef on the version of base to use the correct one.
Okey I have created an issue for this. Will try to fix this at some point of time.
We have introduces handwritten implementations again so as to compare it with that exposed by verse. The output of the benchmarking code is now in csv for better comparison.