On Mon, Feb 06, 2017 at 05:08:17PM -0800, Stephen Boyd wrote: > On 02/06/2017 05:01 PM, Stephen Boyd wrote: > > When I compile files with sparse, I get these sorts of warnings: > > > > arch/arm64/include/asm/lse.h:14:28: warning: Unknown escape 'l' > > arch/arm64/include/asm/lse.h:14:37: warning: Unknown escape 'l' > > arch/arm64/include/asm/alternative.h:172:28: warning: Unknown escape 'o' > > > > This is because sparse is trying to tokenize these files and sees > > a line like this: > > > > alternative_insn "\llsc", "\lse", ARM64_HAS_LSE_ATOMICS > > > > It gets past alternative_insn part and then sees the start of a > > string with the double quote character. So sparse starts to parse > > the string (eat_string() in the sparse code) but the string has > > an escape character '\' in it. Sparse sees the escape character, > > so it checks to see if it's an escape sequence, but '\l' isn't. > > This causes sparse to spit out this warning of an unknown escape > > sequence 'l'. > > > > In reality, sparse isn't going to use these macros anyway because > > this whole thing is inside an __ASSEMBLER__ ifdef. Yes, annoying. Conversion of escaped characters is supposed to be done just after preprocessing. It's definitively a bug. Luc >From 786877f6fa5a3b49c92ef08b28c19e58b75ba8e8 Mon Sep 17 00:00:00 2001 From: Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx> Date: Tue, 7 Feb 2017 02:37:00 +0100 Subject: [PATCH] add testcase for wrong early escape conversion Reported-by: Stephen Boyd <sboyd@xxxxxxxxxxxxxx> Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx> --- validation/preprocessor/early-escape.c | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) create mode 100644 validation/preprocessor/early-escape.c diff --git a/validation/preprocessor/early-escape.c b/validation/preprocessor/early-escape.c new file mode 100644 index 000000000..c7beba5d8 --- /dev/null +++ b/validation/preprocessor/early-escape.c @@ -0,0 +1,23 @@ +#if 0 +"\l" +#endif + +/* + * check-description: + * Following the C standard, escape conversion must be + * done in phase 5, just after preprocessing and just + * before string concatenation. So we're not supposed + * to receive a diagnostic for an unknown escape char + * for a token which is excluded by the preprocessor. + * check-name: early-escape + * check-command: sparse -E $file + * check-known-to-fail + * + * check-output-start + + + * check-output-end + * + * check-error-start + * check-error-end + */ -- 2.11.0 -- To unsubscribe from this list: send the line "unsubscribe linux-sparse" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
