1214037 - MOZ_MUST_USE treats expressions evaluating to lvalue references as MUST_USE

Status: RESOLVED FIXED

(Developer Infrastructure :: Source Code Analysis, defect)

Description

[Nika Layzell [:nika] (ni? for response)]

This means that we have some annoying effects, such as making `a = b` where the type of `a` is MOZ_MUST_USE complain about an unused value (as the assignment expression returns an lvalue reference to the variable).

Another option would be to instead special-case the assignment operator.

Comment 1

[Nika Layzell [:nika] (ni? for response)]

Attachment: Don't consider lvalue references to MOZ_MUST_USE types as MOZ_MUST_USE

Comment 2

[(no longer active)]

Comment on attachment 8673469 [details] [diff] [review]
Don't consider lvalue references to MOZ_MUST_USE types as MOZ_MUST_USE

Review of attachment 8673469 [details] [diff] [review]:
-----------------------------------------------------------------

I think this is a quirk of built-in operator=(), and it's definitely the wrong thing to white-list this for types such as already_AddRefed.  I'd prefer if you special case operator=() assuming that is the only thing that blows up, and Seth's try push seems to suggest that.

Comment 3

[Nika Layzell [:nika] (ni? for response)]

Attachment: Don't consider the result of an assignment expression MOZ_MUST_USE

Comment 4

[Nika Layzell [:nika] (ni? for response)]

a SA run on try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=0df3260d0bf3

Comment 5

[(no longer active)]

Comment on attachment 8673817 [details] [diff] [review]
Don't consider the result of an assignment expression MOZ_MUST_USE

Review of attachment 8673817 [details] [diff] [review]:
-----------------------------------------------------------------

Great, thanks a lot!

Comment 6

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/ab276d264008

Comment 7

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/ab276d264008