Conversation
|
Discussion point: should this be
|
The `reload` request instructs the build server to reload the build configuration. This request should be supported by build tools that keep their state in memory. If the `reload` request returns with an error, it's expected that other requests respond with the previously known "good" state. This request is motivated by tools such as sbt, which maintain an in-memory state of the build, where reloading the build on-demand is not considered practical.
487b1f7 to
efd58b3
Compare
|
I'll make a closer review at this soon, but I think it should be |
True, though the workspace is implicitly provided via the working directory of the build server, so we probably don't need a |
|
I think we do because some build server does not "live" in the same working directory. Bloop, for instance, can manage many workspaces and he wants to know which one to reload. |
Yes, but the server that the clients speaks to is always started in the workspace root directory. Note we don't have any root uri parameter for |
|
Oh yes you are right. Thank you for the clarification. |
|
I am ready to submit the implementation of See sbt/sbt#5838 |
olafurpg
left a comment
There was a problem hiding this comment.
LGTM 👍 I'm open for both build/reload and workspace/reload, the comment is just a nit, feel free to ignore
The `reload` request instructs the build server to reload the build configuration. This request should be supported by build tools that keep their state in memory. If the `reload` request returns with an error, it's expected that other requests respond with the previously known "good" state. This request is motivated by tools such as sbt, which maintain an in-memory state of the build, where reloading the build on-demand is not considered practical.
|
updated to |
The
reloadrequest instructs the build server to reload the build configuration.This request should be supported by build tools that keep their state in memory.
If the
reloadrequest returns with an error, it's expected that other requestsrespond with the previously known "good" state.
This request is motivated by tools such as sbt, which maintain an in-memory state of the build, where reloading the build on-demand is not considered practical.
Prompted by discussion in sbt/sbt#5783
Solves #134
cc @adpi2