wrong protototype for LAPACK_lsame?

[stefanozampini]

Description
both MKL and BLIS define lsame as
lsame( const char* ca, const char* cb...)
while this code uses
same( char* ca, char* cb..

diff --git a/LAPACKE/include/lapack.h b/LAPACKE/include/lapack.h
index a70352acd..efc7068e7 100644
--- a/LAPACKE/include/lapack.h
+++ b/LAPACKE/include/lapack.h
@@ -115,7 +115,7 @@ typedef lapack_logical (*LAPACK_Z_SELECT2)
     ( const lapack_complex_double*, const lapack_complex_double* );
 
 #define LAPACK_lsame LAPACK_GLOBAL(lsame,LSAME)
-lapack_logical LAPACK_lsame( char* ca,  char* cb,
+lapack_logical LAPACK_lsame( const char* ca,  const char* cb,
                               lapack_int lca, lapack_int lcb );

Checklist

• I've included a minimal example to reproduce the issue
• I'd be willing to make a PR to solve this issue

[weslleyspereira]

Hi @stefanozampini. I believe lsame( const char* ca, const char* cb...) is the most appropriate interface. Would you like to propose a PR with the changes?

I looked at OpenBLAS for curiosity. They use lsame( char* ca, char* cb...) - https://github.com/xianyi/OpenBLAS/blob/develop/kernel/generic/lsame.c

[ivan-pi]

For the record, problems encountered by BLIS due to differing const attributes and duplicate inclusion of the prototype are documented here: flame/blis#145 (see posts closte to flame/blis#145 (comment)).

[martin-frbg]

Makes sense to me - the non-const lsame in OpenBLAS mentioned by weslleyspereira above is unchanged from the 2009 GotoBLAS2 and was certainly chosen to match the netlib LAPACK at that time.