feat: add debug logging of repo sub processes #1735
Conversation
Debugging what the repo rules are up to can be difficult because Bazel doesn't provide many facilities to inspect what they're up to. This adds an environment variable, `RULES_PYTHON_REPO_DEBUG`, that, when set, will make our repo rules print out detailed information about the subprocesses they are running. This also makes failed commands dump much more comprehensive information.
rickeylev
changed the title
| args, | ||
| repo_utils.execute_checked( | ||
| rctx, | ||
| op = "whl_library.ResolveRequirement({}, {})".format(rctx.attr.name, rctx.attr.requirement), |
There was a problem hiding this comment.
nit: we could also have this semantics for the op, where the first item is the function name and the remaining are parameters which will be formated to: op[0] + "(" + ",".join(op[1:]) + ")"
| op = "whl_library.ResolveRequirement({}, {})".format(rctx.attr.name, rctx.attr.requirement), | |
| op = ["whl_library.ResolveRequirement", rctx.attr.name, rctx.attr.requirement], |
There was a problem hiding this comment.
I like the idea. It also gave me another idea: automatically add the "whl_library" part (the repo rule name) by looking for a _repo_name attribute we set on all our repo rules.
But, I'm strapped for time today, so I'm going to merge this as-is.
| args, | ||
| repo_utils.execute_checked( | ||
| rctx, | ||
| op = "whl_library.ResolveRequirement({}, {})".format(rctx.attr.name, rctx.attr.requirement), |
There was a problem hiding this comment.
I like the idea. It also gave me another idea: automatically add the "whl_library" part (the repo rule name) by looking for a _repo_name attribute we set on all our repo rules.
But, I'm strapped for time today, so I'm going to merge this as-is.
…on into repos.cmds
Debugging what the repo rules are up to can be difficult because Bazel doesn't provide many facilities to inspect what they're up to. This adds an environment variable,
RULES_PYTHON_REPO_DEBUG, that, when set, will make our repo rules print out detailed information about the subprocesses they are running.This also makes failed commands dump much more comprehensive information.
This was driven by the recent report of failures on Windows during a repo rule.