ghc-mod will always use UTF-8 encoding on output, regardless of the
current encoding. This is a problem for syntastic because it mangles
the filenames. The commit below is a (perhaps misguided) attempt to
fix that, by calling iconv() to re-encode output from ghc-mod to Vim's
encoding.
Undocumented option "bash_hack" is gone.
New function syntastic#util#system(), that sets up the environment and
is now used consistently everywhere.
Unrelated fix for the elixir checker (syntax).
Error sorting can't be done as a postprocess function called from
SyntasticMake(), since the final values of some relevant fields (f.i.
type) might not be known yet at that point. Solution: move sorting to
getLocListRaw(), after per-checker quiet_messages. New checker methods
getWantSort() / setWantSort() are needed.
Second problem: some checkers return screen columns mixed with byte
indices. Solution: compute screen columns as needed. Sadly, everything
about working with screen columns is fragile.
Especially for 'hdevtools' this results into a more robust behaviour,
because 'hdevtools' starts a background process and changing the
current directory doesn't affect the current directory of the background
process.
Optimisation: merge handling of global and per-checker filters.
Feature: allow [] and '' values for quiet_messages filter elements.
Feature: allow overrides for quiet_messages filters.
Feature: buffer-local auto_jump and quiet_messages.
Safety: add magic specifiers to catch regexps.
Cleanup: minor restructuring for the forthcoming foreign checkers
feature.
(1) Checkers now have an _exec attribute, and an accessor getExec().
(2) CreateAndRegisterChecker() initializes _exec from an optional argument
'exec'. If this argument is missing, 'name' is used instead.
(3) Functions SyntaxCheckers_*_IsAvailable() are now dictionary functions.
(4) Functions SyntaxCheckers_*_IsAvailable() are now optional. When
they are missing, they are assumed to return executable(expand(self.getExec())).
(5) Argument 'exe' of function syntastic#makeprg#build() is now optional.
If this argument is missing, expand(self.getExec()) is used to set checker
executables.
Make SyntaxCheckers_*_GetLocList() dictionary functions.
Pass a reference to the current checker to syntastic#makeprg#build().
Add an optional 'redirect' argument to CreateAndRegisterChecker().
Change the sh checker to use the new dictionary functions.
Add a new registry method getLocListRaw() (needed for the sh checker).
Make the "compressWhitespace" filter kill \001 characters.
Make the ghc-mod checker use the "compressWhitespace" filter.
Minor formatting of the hlint checker.
This patch adds an option 'postprocess' to SyntasticMake(). The value
of this option is a list of names. Each name is translated to a
function syntastic#postprocess#name(). These functions are applied in
order to the list of errors just before SyntasticMake() returns. They
take a single parameter, the list of errors, and are supposed to
returned the processed list.
Make all the easy updates. There are still quite a few to do, but in
doing these ones I can see that syntastic#makeprg#build() needs to
accept a few more options. Namely:
* "postargs" that appear after the filename
* "tail" that appears after everything - used for things like
redirecting output and piping to grep/sed/etc
* the filename itself - only the java checkers needed this since they
specify the directory of the file to check as well
There are still a few other things to do as well:
* remove the options from the checkers that are now provided by
syntastic#makeprg#build implicitly - i.e. the checker exe and args.
* also, we need to doc the above implicit checker options
Since we're only using ghc to check and lint the file and not actually create output, we can pass -fno-code which omits code generation and speeds up the check quite a bit.
