Unintuitive column number when source file contains UTF-8 BOM

[Quuxplusone]

Bugzilla Link	PR21144
Status	NEW
Importance	P normal
Reported by	xda@cypress.com
Reported on	2014-10-02 17:18:49 -0700
Last modified on	2017-06-01 03:08:49 -0700
Version	3.7
Hardware	PC Windows NT
CC	Erik.Verbruggen@Me.com, ivan.donchevskii@qt.io, llvm-bugs@lists.llvm.org
Fixed by commit(s)
Attachments	utf8.zip (289 bytes, application/octet-stream) main.cpp (90 bytes, text/plain) main.cpp (91 bytes, text/plain)
Blocks
Blocked by
See also

Created attachment 13127
UTF-8 and Latin-1 source files

When there is a diagnostic on the first line of a file with a UTF-8 BOM, the
column number is 3 characters greater than expected:

C:\Program Files (x86)\LLVM\bin>clang -W -fsyntax-only
C:\Users\xda\Desktop\utf8.c
C:\Users\xda\Desktop\utf8.c:1:5: error: "whoops!"
<U+FEFF>#error "whoops!"
         ^
1 error generated.

C:\Program Files (x86)\LLVM\bin>clang -W -fsyntax-only
C:\Users\xda\Desktop\lat1.c
C:\Users\xda\Desktop\lat1.c:1:2: error: "whoops!"
#error "whoops!"
 ^
1 error generated.

[Quuxplusone]

Attached utf8.zip (289 bytes, application/octet-stream): UTF-8 and Latin-1 source files

[Quuxplusone]

Attached main.cpp (90 bytes, text/plain): wrong column number

[Quuxplusone]

Attached main.cpp (91 bytes, text/plain): file with broken diagnostics