On Wed, Jul 7, 2010 at 4:02 PM, Stefan Reinauer stefan.reinauer@coresystems.de wrote:
On 7/7/10 8:28 PM, Myles Watson wrote:
On Wed, Jul 7, 2010 at 8:24 AM, Myles Watson mylesgw@gmail.com wrote:
Rev 5653.
Thanks, Myles
Not sure why there were no errors, but:
9 src/include/pc80/mc146818rtc.h:89:26: warning: option_table.h: No such file or directory
That's odd. I don't see that warning. Does it have to do with parallel compilation and insufficient dependency checking? Which boards have the warning?
http://qa.coreboot.org/log_buildbrd.php?revision=5657&device=dbm690t&...
option_table.h doesn't get generated until way down the list (after the warnings) It looks like the first board that needs this:
HOSTCC coreboot-builds/sharedutils/options/build_opt_tbl
And the rest of the build doesn't wait for it.
This sure looks like a race... question is, what do we do about it? Force an early build of option_table.h outside of the normal dependency system?
I don't know what the best fix is. The problem is the building of the tool, so abuild could just build the tool before building any boards. We could also create option_table.h as part of the configure process.
Why does gcc only print a warning instead of an error here? Are those #includes not needed at all?
USE_OPTION_TABLE defaults to false. HAVE_OPTION_TABLE triggers the #include so that the options get defined. If you select USE_OPTION_TABLE and create the race condition, gcc will error out.
Steps to exercise the race condition for the curious:
rm -rf build make oldconfig make -j4
Thanks, Myles