WIP: Parse endpoints using argparse #76
Conversation
|
Happy to re-add |
|
This is really coincidental; I've run into this issue also the last few days, and just last night (shortly before this issue), I opened a PR to fix this as well #74 I take a different approach though, just adding a patter to match window's drive letters and ignore, similar to how it already did for |
|
|
||
| args, unknown = parser.parse_known_args(argv[:endpoint_index]) |
There was a problem hiding this comment.
I've found a regression caused by this. If the program args contains hyphenated args, e.g. ["com.pinterest:ktlint", "-F", "/c/path/to/file.kt"] , the hyphenated arg becomes a JVM arg.
There was a problem hiding this comment.
A workaround for this is for people to use -- , i.e. ["com.pinterest:ktlint", "--", "-F", "/c/path/to/file.kt"]
Requiring that workaround would mean this is a breaking change.
Looking for other solutions ...
There was a problem hiding this comment.
Using -- isnt enough.
["--add-opens", "java.base/java.lang=ALL-UNNAMED", "com.pinterest:ktlint", "--", "-F", "/c/path/to/file.kt"] ends up with jvm arg ["--add-opens"] and endpoint java.base/java.lang=ALL-UNNAMED. :-(
|
I think there are benefits in where I was going with this, but I want to first get tests in place for the current behaviour, before attempting to take this approach further. I'll get a WIP up shortly. |
additional endpoints and shortcuts fail
jayvdb
force-pushed
the
argparse_program_args
branch
from
6819b3a
to
133601a
Compare
jayvdb
changed the title
|
@ctrueden, I am pretty confident that I can get this working correctly, resulting in no more attempt to parse the program_args and all the problems that can cause. OTOH, complexity goes through the roof, and we'd be using 'private' parts of the Should I keep going? |
No description provided.