Comment 1:

This looks like a great error. What's the problem?

Comment 2:

M is also a method, gri@ tested it on ggcgo and it compiles.
So this is either a gc bug or a gccgo one, in any case it seems there's a discrepancy

Comment 3:

Based on the current spec, I believe this is a bug in gccgo.  The spec says that the
method set of a struct includes all promoted methods, not all methods.  A promoted
method is a method M such that v.M denotes that method.  In this case v.M denotes the
embedded field, not the method.  So M should not be in the method set, and the
assignment should fail.
Of course we could decide that the spec is wrong, but I think the spec is plausible, and
it follows that gccgo is in error.
It would be great if you could add this test to the testsuite.

Labels changed: added priority-later, gccgo, removed priority-triage.

Comment 4:

I updated the summary to better match the problem based on rsc and gri comments.
How should I add this test to the testsuite? Also, if I add it it will fail and
according to the contribution guidelines I can't add a patch that introduces new errors,
therefore I'd need to fix the problem first. And I'm not sure it's worth spending time
for me going through the gccgo code at this time.

Comment 5:

Adding a test case to the gc testsuite will not cause the gccgo testsuite to fail.

Comment 6:

Labels changed: added go1.2, removed go1.1.

Comment 7:

Removing Go1.2 label, as these are not blocking the release.

Labels changed: removed go1.2.

Comment 8:

Labels changed: added go1.3maybe.

Comment 9:

Labels changed: added release-none, removed go1.3maybe.

Comment 10:

Labels changed: added repo-gccgo.

Comment 11:

Fixed on tip.
https://golang.org/cl/147280043/

Status changed to Fixed.

Comment 12:

CL https://golang.org/cl/151410043 mentions this issue.