On 4/8/19 1:09 AM, Andreas Schwab wrote:
FAIL: gcc.target/riscv/arch-1.c (test for errors, line ) FAIL: gcc.target/riscv/arch-1.c (test for excess errors) Excess errors: cc1: error: '-march=rv32I': first ISA subset must be 'e', 'i' or 'g'
The testcase has to be fixed, but there is a locale issue that has me confused at the moment, because this testcase is actually working on one of my machines and failing on another one. Looking at them, I see that one is printing a backquote and one is printing a regular single quote. I had thought that the gcc testsuite set the locale environment variables to avoid this problem, but maybe there is a problem with the riscv.exp file not including something. I've also seen this testcase fail before Martin's patch went in, so this isn't a new problem, just shifted around a bit. Since this is apparently just a testsuite problem, it isn't at the top of my todo list at the moment. Or maybe it is a termcap problem not a locale problem. Or maybe the right fix is to just use a regexp to match any quote character.
Jim
