cvs commit: src/share/man/man4 Makefile src/sys/dev/cxgb cxgb_adapter.h cxgb_config.h cxgb_ioctl.h cxgb_lro.c cxgb_main.c cxgb_osdep.h cxgb_sge.c t3fw-3.2.bin.gz.uu src/sys/dev/cxgb/common cxgb_ael1002.c cxgb_common.h cxgb_firmware_exports.h cxgb

[Peter Pentchev]

On Thu, Mar 15, 2007 at 10:35:49PM -0700, Kip Macy wrote:
> 
> 
> On 3/15/07, Peter Pentchev <roam at ringlet.net> wrote:
> >On Thu, 15 Mar 2007 at 03:06:32 +0000 (UTC), Kip Macy wrote:
> >>
> >> kmacy       2007-03-15 03:06:32 UTC
> >>
> >>   FreeBSD src repository
> >>
> >>   Modified files:        (Branch: RELENG_6)
> >>     share/man/man4       Makefile
> >>   Added files:           (Branch: RELENG_6)
> >>     sys/dev/cxgb         cxgb_adapter.h cxgb_config.h cxgb_ioctl.h
> >>                          cxgb_lro.c cxgb_main.c cxgb_osdep.h
> >>                          cxgb_sge.c t3fw-3.2.bin.gz.uu
> >>     sys/dev/cxgb/common  cxgb_ael1002.c cxgb_common.h
> >>                          cxgb_firmware_exports.h cxgb_mc5.c
> >>                          cxgb_mv88e1xxx.c cxgb_regs.h
> >>                          cxgb_sge_defs.h cxgb_t3_cpl.h
> >>                          cxgb_t3_hw.c cxgb_tcb.h cxgb_version.h
> >>                          cxgb_vsc8211.c cxgb_xgmac.c
> >>     sys/modules/cxgb     Makefile
> >>   Log:
> >>   MFC Chelsio T3 10 Gigabit Ethernet support
> >>
> >>   Don't hook into build just
> >
> >Is it possible that the GCC version in 6.x-STABLE is different from
> >that in -CURRENT?  On my laptop (i386, yesterday's -STABLE) LINT failed
> >to build with the following:
> >
> >/fs/6/usr.src/sys/dev/cxgb/cxgb_sge.c: In function `sge_timer_reclaim':
> >/fs/6/usr.src/sys/dev/cxgb/cxgb_sge.c:172: warning: inlining failed in 
> >call to 'refill_rspq': function body not available
> >/fs/6/usr.src/sys/dev/cxgb/cxgb_sge.c:699: warning: called from here
> >
> >It seems it simply needed refill_rspq() to be moved up, so that
> >the compiler could actually see the function body before first use;
> >or is this a bug in GCC?
> >
> >Anyway, here's a simple patch that fixed the LINT build for me.
> 
> I'm not saying it isn't a problem - but I can't reproduce it on my
> system. I see all sorts of inlining complaints from other modules but
> not cxgb. Are you sure you don't have a modified make.conf?

Actually, yes, I was using a slightly modified make.conf.  After playing
around for a while it turned out that the "problem" was a CONFIGARGS?=-g
line, which changed the gcc invocation to -O instead of the default -O2.

Again, after playing around with the gcc options a bit, it turns out
that -O2 includes -funit-at-a-time, making gcc parse the whole cxgb_sge.c
before the compilation, so that it can see the refill_rspq() function
body and it can inline it properly.

So... I'm not sure what the proper solution ought to be - either figure
out a way to add -funit-at-a-time to CFLAGS (which might not work all
that well with non-gcc compilers, I guess), or move the function body up
as per my original suggestion.

G'luck,
Peter

-- 
Peter Pentchev	roam at ringlet.net    roam at cnsys.bg    roam at FreeBSD.org
PGP key:	http://people.FreeBSD.org/~roam/roam.key.asc
Key fingerprint	FDBA FD79 C26F 3C51 C95E  DF9E ED18 B68D 1619 4553
I had to translate this sentence into English because I could not read the original Sanskrit.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/cvs-src/attachments/20070316/9c79c348/attachment.pgp