Stream: t-compiler/wg-meta

Topic: mcp process and user-facing changes


nikomatsakis (May 07 2020 at 15:45, on Zulip):

This was I think in relation to our discussion around "what is the procedure to add a new flag to the compiler"

nikomatsakis (May 07 2020 at 15:45, on Zulip):

I think the goal of the MCP process was more around "big changes to the code" than "all decisions"

nikomatsakis (May 07 2020 at 15:46, on Zulip):

But we did also discuss the idea of using it for command-line flags

nikomatsakis (May 07 2020 at 15:46, on Zulip):

I don't actually think that submitting a PR has worked that well in the past -- it works ok for folks "in the know", but I think that for folks coming from outside, it's not very clear what the expectations are (e.g., I get pings about this from time to time)

nikomatsakis (May 07 2020 at 15:47, on Zulip):

Plus I think we don't always have guidance for writing reports and the like

nikomatsakis (May 07 2020 at 15:50, on Zulip):

I guess this fits into the question of "so you want to do X" in the compiler, trying to collect that information

nikomatsakis (May 07 2020 at 16:04, on Zulip):

so here is one possible workflow--

nikomatsakis (May 07 2020 at 16:04, on Zulip):

https://hackmd.io/Ncw1xtSnQCKsx4cFgFz1Ow?both

nikomatsakis (May 07 2020 at 16:04, on Zulip):

but the other possibility is just to say "don't use MCP for outward facing changes"

nikomatsakis (May 07 2020 at 16:05, on Zulip):

@pnkfelix I'm curious to get your opinion

nikomatsakis (May 07 2020 at 16:05, on Zulip):

one obvious thing: it'd be nice to move all this workflow to one bot, e.g., rfcbot could have a "second" command

nikomatsakis (May 07 2020 at 16:05, on Zulip):

that is the same as rfcbot fcp merge except it doesn't require all the boxes to be explicitly checked

nikomatsakis (May 07 2020 at 16:06, on Zulip):

(I remember @pnkfelix you were advocating for "one process")

nikomatsakis (May 07 2020 at 16:10, on Zulip):

ok well I edited the RFC and kind of just leftit a bit vague, saying that the second process isn't enough to approve outward facing changes

nikomatsakis (May 07 2020 at 16:10, on Zulip):

but I think we should come back to this

Last update: Sep 28 2020 at 16:45UTC