May also fix#1154, #1033, #708, #536, #73: testing is needed.
This aims to solve all problems where misconfigured locales lead to
parsers failing on utf8-encoded data. This should hopefully avoid
encoding issues, but since it fundamentally alters how encoding is dealt
with it may lead to unexpected outcomes. Widespread testing on a number
of different platforms would be useful.
env -S isn't a thing on linux of course. Go back to using standard
env, which means using a stack options line, which means not using
"ghc". This new setup is probably simpler anyway. I've just had to
give up on the goal of having each script's required packages being
defined in one place; now (to they extent they are required) they
must be defined both in the script header and in compile.sh.
Using stack's script command meant that the scripts needed to be
compatible, and regularly tested, with a hledger release in stackage,
rather than the latest hledger source. This created hassles for
maintainers, contributors and sometimes for users.
To simplify things overall, we now require script users to check out
the hledger source tree and run the scripts (or, bin/compile.sh) from
there once so they compile themselves. Some notes on alternative
setups are included (in one of the scripts, and referenced by the
others). This ensures that users and our CI tests are building scripts
the same way.
Current stack does not allow a stack options line to be used with the
"stack ghc" command, unfortunately, so instead we are using env's -S
flag, which hopefully has sufficiently wide support by now, and
putting all arguments in the shebang line.
This method will probably require complete explicit --package options,
unlike "stack script", so more testing and tweaking is expected.
Probably we're going to end up with some long shebang lines.
This isn't pretty but seems like a possible way to keep things
manageable.
