what is cooking in sparse-next-20170805

what is cooking in sparse-next-20170805

[Christopher Li]

Two new series from Luc has been applied.
- fix: give a type to bad conditionnal expressions
  I confirm the wine source file checking went from 23s to 3.x seconds.

- fix list corruption with recursive remove_usage()
   this replace 7d54acb0d6d4a9dff3c91688b0223cd4c4a6d6ea
   "Let pseudo->users loop on duplicate versin of list"

The kernel compile check shows that has output difference in 3 files
compare to rc4. I include the diff here. I will dig a bit more which change
cause it. It seems reasonable enough.  There is one context imbalance
warning was gone. My guess is that the conditional expression patch
make sparse understand the source code better now.

This should be very very close the RC5 release.

Go give it more test!

Chris




$ diff linux-checker/rc4 linux-checker/exp -ruN
diff -ruN linux-checker/rc4/arch/x86/platform/uv/uv_nmi.sp
linux-checker/exp/arch/x86/platform/uv/uv_nmi.sp
--- linux-checker/rc4/arch/x86/platform/uv/uv_nmi.sp 2017-08-05
05:07:52.176462385 -0700
+++ linux-checker/exp/arch/x86/platform/uv/uv_nmi.sp 2017-08-05
04:56:42.959747891 -0700
@@ -24,5 +24,5 @@
 ./arch/x86/include/asm/uv/uv_hub.h:678:41:    got unsigned long *
 arch/x86/platform/uv/uv_nmi.c:908:5: warning: symbol 'uv_handle_nmi'
was not declared. Should it be static?
 arch/x86/platform/uv/uv_nmi.c:1016:13: warning: symbol
'uv_nmi_setup_common' was not declared. Should it be static?
-arch/x86/platform/uv/uv_nmi.c:482:12: warning: context imbalance in
'uv_check_nmi' - different lock contexts for basic block
+arch/x86/platform/uv/uv_nmi.c:541:9: warning: context imbalance in
'uv_check_nmi' - different lock contexts for basic block
 arch/x86/platform/uv/uv_nmi.c:559:17: warning: context imbalance in
'uv_handle_nmi' - unexpected unlock
diff -ruN linux-checker/rc4/drivers/auxdisplay/panel.sp
linux-checker/exp/drivers/auxdisplay/panel.sp
--- linux-checker/rc4/drivers/auxdisplay/panel.sp 2017-08-05
05:07:54.558438024 -0700
+++ linux-checker/exp/drivers/auxdisplay/panel.sp 2017-08-05
04:56:45.285724297 -0700
@@ -1 +0,0 @@
-drivers/auxdisplay/panel.c:1402:17: warning: context imbalance in
'panel_scan_timer' - different lock contexts for basic block
diff -ruN linux-checker/rc4/drivers/net/ethernet/chelsio/cxgb3/sge.sp
linux-checker/exp/drivers/net/ethernet/chelsio/cxgb3/sge.sp
--- linux-checker/rc4/drivers/net/ethernet/chelsio/cxgb3/sge.sp
2017-08-05 05:08:45.045911820 -0700
+++ linux-checker/exp/drivers/net/ethernet/chelsio/cxgb3/sge.sp
2017-08-05 04:57:32.503081874 -0700
@@ -13,4 +13,3 @@
 drivers/net/ethernet/chelsio/cxgb3/sge.c:2371:47: warning: incorrect
type in assignment (different base types)
 drivers/net/ethernet/chelsio/cxgb3/sge.c:2371:47:    expected
unsigned int [unsigned] [usertype] priority
 drivers/net/ethernet/chelsio/cxgb3/sge.c:2371:47:    got restricted
__be32 [assigned] [usertype] rss_lo
-drivers/net/ethernet/chelsio/cxgb3/sge.c:2936:9: warning: context
imbalance in 'sge_timer_rx' - different lock contexts for basic block
diff -ruN linux-checker/rc4/drivers/staging/lustre/lustre/obdclass/lu_object.sp
linux-checker/exp/drivers/staging/lustre/lustre/obdclass/lu_object.sp
--- linux-checker/rc4/drivers/staging/lustre/lustre/obdclass/lu_object.sp
2017-08-05 05:09:19.649552223 -0700
+++ linux-checker/exp/drivers/staging/lustre/lustre/obdclass/lu_object.sp
2017-08-05 04:58:06.379516299 -0700
@@ -1,2 +1,2 @@
-drivers/staging/lustre/lustre/obdclass/lu_object.c:95:6: warning:
context imbalance in 'lu_object_put' - different lock contexts for
basic block
+drivers/staging/lustre/lustre/obdclass/lu_object.c:95:6: warning:
context imbalance in 'lu_object_put' - wrong count at exit
 drivers/staging/lustre/lustre/obdclass/lu_object.c:1727:54: warning:
context imbalance in 'lu_context_exit' - different lock contexts for
basic block

Re: what is cooking in sparse-next-20170805

[Christopher Li]

On Sat, Aug 5, 2017 at 8:42 AM, Christopher Li <sparse@chrisli.org> wrote:
> diff -ruN linux-checker/rc4/drivers/auxdisplay/panel.sp
> linux-checker/exp/drivers/auxdisplay/panel.sp
> --- linux-checker/rc4/drivers/auxdisplay/panel.sp 2017-08-05
> 05:07:54.558438024 -0700
> +++ linux-checker/exp/drivers/auxdisplay/panel.sp 2017-08-05
> 04:56:45.285724297 -0700
> @@ -1 +0,0 @@
> -drivers/auxdisplay/panel.c:1402:17: warning: context imbalance in
> 'panel_scan_timer' - different lock contexts for basic block
> diff -ruN linux-checker/rc4/drivers/net/ethernet/chelsio/cxgb3/sge.sp
> linux-checker/exp/drivers/net/ethernet/chelsio/cxgb3/sge.sp
> --- linux-checker/rc4/drivers/net/ethernet/chelsio/cxgb3/sge.sp
> 2017-08-05 05:08:45.045911820 -0700
> +++ linux-checker/exp/drivers/net/ethernet/chelsio/cxgb3/sge.sp
> 2017-08-05 04:57:32.503081874 -0700
> @@ -13,4 +13,3 @@
>  drivers/net/ethernet/chelsio/cxgb3/sge.c:2371:47: warning: incorrect
> type in assignment (different base types)
>  drivers/net/ethernet/chelsio/cxgb3/sge.c:2371:47:    expected
> unsigned int [unsigned] [usertype] priority
>  drivers/net/ethernet/chelsio/cxgb3/sge.c:2371:47:    got restricted
> __be32 [assigned] [usertype] rss_lo
> -drivers/net/ethernet/chelsio/cxgb3/sge.c:2936:9: warning: context
> imbalance in 'sge_timer_rx' - different lock contexts for basic block

This different is actually trace back to this change, not by the recent
two series:

c7dd24480716c2bf853b63d462c3c8be84b78359
fix BB dependencies on phi-nodes

I some how did not catch it earlier.

Chris

Re: what is cooking in sparse-next-20170805

[Dibyendu Majumdar]

Hi Chris,

Does sparse-next now have all the changes/fixes since January? I am
still waiting for this to stabilize before attempting a merge into my
project.

Thanks and Regards
Dibyendu

Re: what is cooking in sparse-next-20170805

[Christopher Li]

On Sat, Aug 5, 2017 at 9:56 AM, Dibyendu Majumdar
<mobile@majumdar.org.uk> wrote:
> Hi Chris,
>
> Does sparse-next now have all the changes/fixes since January? I am
> still waiting for this to stabilize before attempting a merge into my
> project.

No. The master branch is the stable point. The sparse-next is playground
for testing patches before merge to master. The history in sparse-next
can get rewrite to remove some temporary fix.

For all the patches, especially the llvm one you are waiting for.
Have to wait for Luc resend the merge request after the release.
This release is just get some stable bits out first. Serve as anchor
point for the coming merges.

Right now we don't have blocker or know to be broken issue
on sparse. The release should happen relative soon.
Famous last word.


Chris

Re: what is cooking in sparse-next-20170805

[Luc Van Oostenryck]

On Sat, Aug 05, 2017 at 08:42:12AM -0400, Christopher Li wrote:
> Two new series from Luc has been applied.
> - fix: give a type to bad conditionnal expressions
>   I confirm the wine source file checking went from 23s to 3.x seconds.
> 
> - fix list corruption with recursive remove_usage()
>    this replace 7d54acb0d6d4a9dff3c91688b0223cd4c4a6d6ea
>    "Let pseudo->users loop on duplicate versin of list"
> 
> The kernel compile check shows that has output difference in 3 files
> compare to rc4. I include the diff here. I will dig a bit more which change
> cause it. It seems reasonable enough.  There is one context imbalance
> warning was gone. My guess is that the conditional expression patch
> make sparse understand the source code better now.

Yes.

Also the context checking warnings are quite sensitive to
small changes in the optimization (via the merging of BBs).
But then it's just a change in the file:line:pos, like in
the first file here under. The other two are different and
I also think they are due to the cond-expr change.

-- Luc