1225642 - Add -Wno-parentheses-equality back to Clang builds that need it

Status: RESOLVED FIXED

(NSS :: Build, defect)

Description

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

Bug 1211568 seems to have slightly broken the OS X build (at least with the compiler version that calls itself "Apple LLVM version 7.0.0 (clang-700.1.76)") while refactoring the -Werror handling, by adding -Wno-parentheses-equality to the wrong make variable.

Comment 1

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

On further thought, this probably affects Clang on other platforms.

Comment 2

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

Attachment: One-line patch

Comment 3

[Martin Thomson [:mt:]]

Comment on attachment 8688670 [details] [diff] [review]
One-line patch

Review of attachment 8688670 [details] [diff] [review]:
-----------------------------------------------------------------

https://hg.mozilla.org/projects/nss/rev/6a631abbf25d

Comment 4

[Martin Thomson [:mt:]]

Jed, do you think that this fix needs to be uplifted to NSS 3.21?  (Keeping in mind that I don't know how we might do that and that we might want to release 3.22 soon-ish anyway.)

Comment 5

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

(In reply to Martin Thomson [:mt:] from comment #4)
> Jed, do you think that this fix needs to be uplifted to NSS 3.21?  (Keeping
> in mind that I don't know how we might do that and that we might want to
> release 3.22 soon-ish anyway.)

…maybe.  NSS_3_21_RTM fails to build for me on OS X 10.10, which is bad, but the other part of this (which I just noticed) is that I'm using ccache.  There isn't an expression of the form ((e1 == e2)) in the original source, but there are some after preprocessing expands macros; clang has enough information to suppress the warning in that case if it's compiling normally, but ccache runs preprocessing separately.  So this wouldn't affect people not using ccache.

But it should be low risk for uplifting if 3.21.1 is going to happen.

Comment 6

[Martin Thomson [:mt:]]

Yes, I discovered that ccache has some unfortunate interactions with clang.  I have turned it off as a result.  It took be a while to discover that Fedora rewrites your path when ccache is installed so that it intercepts calls to clang/gcc/cc.

I'll try to remember this if we need a 3.21.1.  My hope is to release 3.22 pretty soon in any case.