865396 - jsd_xpc.cpp:47:61: warning: invalid suffix on literal; C++11 requires a space between literal and identifier [-Wliteral-suffix], for "+++++ "name

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Daniel Holbert [:dholbert]]

Build warnings for jsd_xpc.cpp in GCC 4.8:

/mozilla/js/jsd/jsd_xpc.cpp:47:61: warning: invalid suffix on literal; C++11 requires a space between literal and identifier [-Wliteral-suffix]
 #   define DEBUG_CREATE(name, count) {count++; DEBUG_COUNT ("+++++ "name,count)}
                                                             ^
/mozilla/js/jsd/jsd_xpc.cpp:48:62: warning: invalid suffix on literal; C++11 requires a space between literal and identifier [-Wliteral-suffix]
 #   define DEBUG_DESTROY(name, count) {count--; DEBUG_COUNT ("----- "name,count)}

MXR link:
 http://mxr.mozilla.org/mozilla-central/source/js/jsd/jsd_xpc.cpp#47

The issue is with
  "+++++ "name
and
  "----- "name

tl;dr, we just need to insert a space before "name" there.

LONGER EXPLANATION:

So, this code is trying to do string concatenation (where 'name' is a macro-parameter for a string literal).  BUT in C++ 11, the code "+++ "name could be interpreted differently, because C++11 has special syntax for characters immediately following a string literal. They're interpreted as being a user-defined literal suffix, sort of like the suffixes after numeric literals in 7.0f and 1000u.

More information here:
  http://en.wikipedia.org/wiki/C%2B%2B11#User-defined_literals
and see in particular this example from that wikipedia page, demonstrating the syntax:
>  OutputType some_variable =   "1234"_ssuffix; // Uses the 'const char *' overload.

SO ANYWAY: If we insert a space between the two string literals, the result is unchanged and it becomes unambiguous even in light of the new C++11 syntax.  Patch coming up to do that.

Comment 1

[Daniel Holbert [:dholbert]]

Attachment: fix (insert space before 'name')

Comment 2

[Daniel Holbert [:dholbert]]

(Note that technically, only suffixes that start with an underscore can have special meaning after a string literal in C++11. But non-underscore-starting suffixes are still reserved for use in *future* specs, according to this reddit comment: http://www.reddit.com/r/cpp/comments/mol07/example_of_c11_compiletime_binary_number_literal/c32m9iq  So that means the 'name' suffix in this bug's flagged lines could feasibly obtain some unwanted special meaning in a future C++ draft.)

Comment 3

[Daniel Holbert [:dholbert]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/d3e73cfeb94c

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/d3e73cfeb94c