2007-05-22 11:51:42

by Bob Tracy

[permalink] [raw]
Subject: [PATCH] aic7xxx/aicasm build failure w/gcc-3.4.6

Second try: originally reported this back on April 17th. 2.6.X
kernel builds started failing after I upgraded my compiler from
gcc-3.3.X to gcc-3.4.6:

make -C drivers/scsi/aic7xxx/aicasm
(...)
gcc -I/usr/include -I. aicasm.c aicasm_symbol.c aicasm_gram.c aicasm_macro_gram.c aicasm_scan.c aicasm_macro_scan.c -o aicasm -ldb
aicasm_gram.y:1948: error: conflicting types for 'yyerror'
aicasm_gram.tab.c:3004: error: previous implicit declaration of 'yyerror' was here
aicasm_macro_gram.y:162: error: conflicting types for 'mmerror'
aicasm_macro_gram.tab.c:1196: error: previous implicit declaration of 'mmerror' was here

As a workaround, commenting out or deleting the "void" declarations
for yyerror() and mmerror() in the respective ".y" files fixes the
problem. A patch to illustrate the offending code is attached, but
there's no "signed-off by" line because I'm certain the final form of
the patch will be different. The patch applies cleanly to at least
2.6.21 and later kernels. gcc-3.3 may have been warning about the
type conflicts, but I didn't notice: gcc-3.4 treats the type conflicts
as errors, so I *did* notice :-).

Here's the "gcc -v" output:

Reading specs from /usr/lib/gcc/i486-slackware-linux/3.4.6/specs
Configured with: ../gcc-3.4.6/configure --prefix=/usr --enable-shared --enable-threads=posix --enable-__cxa_atexit --disable-checking --with-gnu-ld --verbose --target=i486-slackware-linux --host=i486-slackware-linux
Thread model: posix
gcc version 3.4.6

--
-----------------------------------------------------------------------
Bob Tracy WTO + WIPO = DMCA? http://www.anti-dmca.org
[email protected]
-----------------------------------------------------------------------


Attachments:
patch22_aicasm (626.00 B)

2007-05-22 14:26:35

by James Bottomley

[permalink] [raw]
Subject: Re: [PATCH] aic7xxx/aicasm build failure w/gcc-3.4.6

On Tue, 2007-05-22 at 06:51 -0500, Bob Tracy wrote:
> Second try: originally reported this back on April 17th. 2.6.X
> kernel builds started failing after I upgraded my compiler from
> gcc-3.3.X to gcc-3.4.6:
>
> make -C drivers/scsi/aic7xxx/aicasm
> (...)
> gcc -I/usr/include -I. aicasm.c aicasm_symbol.c aicasm_gram.c aicasm_macro_gram.c aicasm_scan.c aicasm_macro_scan.c -o aicasm -ldb
> aicasm_gram.y:1948: error: conflicting types for 'yyerror'
> aicasm_gram.tab.c:3004: error: previous implicit declaration of 'yyerror' was here
> aicasm_macro_gram.y:162: error: conflicting types for 'mmerror'
> aicasm_macro_gram.tab.c:1196: error: previous implicit declaration of 'mmerror' was here
>
> As a workaround, commenting out or deleting the "void" declarations
> for yyerror() and mmerror() in the respective ".y" files fixes the
> problem.

We really don't want gcc making assumptions about prototypes ... even if
it's getting them right in all likelihood (doubtless unprototyped
assumed functions will become a warning and then an error in later gcc
versions ...), so this is a better fix

James

diff --git a/drivers/scsi/aic7xxx/aicasm/aicasm_gram.y b/drivers/scsi/aic7xxx/aicasm/aicasm_gram.y
index c328596..6066998 100644
--- a/drivers/scsi/aic7xxx/aicasm/aicasm_gram.y
+++ b/drivers/scsi/aic7xxx/aicasm/aicasm_gram.y
@@ -106,6 +106,7 @@ static void make_expression(expression_t *immed, int value);
static void add_conditional(symbol_t *symbol);
static void add_version(const char *verstring);
static int is_download_const(expression_t *immed);
+void yyerror(const char *string);

#define SRAM_SYMNAME "SRAM_BASE"
#define SCB_SYMNAME "SCB_BASE"
diff --git a/drivers/scsi/aic7xxx/aicasm/aicasm_macro_gram.y b/drivers/scsi/aic7xxx/aicasm/aicasm_macro_gram.y
index 439f760..ff46aa6 100644
--- a/drivers/scsi/aic7xxx/aicasm/aicasm_macro_gram.y
+++ b/drivers/scsi/aic7xxx/aicasm/aicasm_macro_gram.y
@@ -65,6 +65,7 @@
static symbol_t *macro_symbol;

static void add_macro_arg(const char *argtext, int position);
+void mmerror(const char *string);

%}



2007-05-22 15:12:00

by Bob Tracy

[permalink] [raw]
Subject: Re: [PATCH] aic7xxx/aicasm build failure w/gcc-3.4.6

James Bottomley wrote:
> We really don't want gcc making assumptions about prototypes ... even if
> it's getting them right in all likelihood (doubtless unprototyped
> assumed functions will become a warning and then an error in later gcc
> versions ...), so this is a better fix

ACK. The fix works here. If you would be so kind, please push it
upstream at your convenience.

gcc-4.X violates the principle of least astonishment over even more
nitnoid matters, but that's another flame for another day.

--
-----------------------------------------------------------------------
Bob Tracy | "Eagles may soar, but weasels don't get
[email protected] | sucked into jet engines." --Anon
-----------------------------------------------------------------------