master branch for upcoming RC5

master branch for upcoming RC5

[Christopher Li]

Just let every know that I already delete all the sparse-next
branch for now.

Master is the the integration branch now, again.

I have add two documents and even the inspect gtk3 fix.

I think the master has every thing I want to have for RC5.
- the wine bug is nail down for good. Thanks Luc for the hard
   work there.
- The ptrlist delete while looping is checked and the offender
   has been change to mark delete rather than actual delete.
- The document and project ideas were added. Thanks Randy
   for correcting my horrible grammar mistakes.

We still have some clouds on the improper SSA forms, I am
confident that I have pretty good ideas what needs to be done
for them in the next release. Luc even have some patches.

Luc, I think the you have some thing to add to as the finial touch.
It is on you now.

Chris

Re: master branch for upcoming RC5

[Luc Van Oostenryck]

On Fri, Aug 11, 2017 at 3:49 PM, Christopher Li <sparse@chrisli.org> wrote:
>
> Luc, I think the you have some thing to add to as the finial touch.
> It is on you now.

Yes, sorry, I haven't been able to send a proper pull request yesterday.
It's on the way now.

--Luc

Re: master branch for upcoming RC5

[Dibyendu Majumdar]

Hi Chris,

On 11 August 2017 at 14:49, Christopher Li <sparse@chrisli.org> wrote:
> - the wine bug is nail down for good. Thanks Luc for the hard
>    work there.
> - The ptrlist delete while looping is checked and the offender
>    has been change to mark delete rather than actual delete.

Were these two issues caused by recent changes - or are they bugs in
the released version of Sparse?
I ran the test case kill-unreachable-phi.c in my repository and did
not see an infinite loop. My repository is way behind the latest
changes in Sparse as I last synced up earlier this year so maybe I am
not seeing this issue?

Regards
Dibyendu

Re: master branch for upcoming RC5

[Luc Van Oostenryck]

On Fri, Aug 11, 2017 at 5:00 PM, Dibyendu Majumdar
<mobile@majumdar.org.uk> wrote:
>
> Were these two issues caused by recent changes - or are they bugs in
> the released version of Sparse?

Both have been there since essentially forever.
It's only recently that we discovered how bad the situation was

> I ran the test case kill-unreachable-phi.c in my repository and did
> not see an infinite loop. My repository is way behind the latest
> changes in Sparse as I last synced up earlier this year so maybe I am
> not seeing this issue?

It's possible. It's the kind of things that is very sensitive to
details like the exact ordering of different operations.

It's probably another change somewhere that triggered a chain
and finally, this bug showed up.

-- Luc

Re: master branch for upcoming RC5

[Christopher Li]

On Fri, Aug 11, 2017 at 11:00 AM, Dibyendu Majumdar
<mobile@majumdar.org.uk> wrote:
> Were these two issues caused by recent changes - or are they bugs in
> the released version of Sparse?

The bug is there in the v0.5.0 version of sparse but not trigger-able.
That is until Luc submit some patch kill dead bb and dead instructions
which make the bug trigger-able.

There is a revert on sparse-next to temporary stop the bug, by just
revert some change that Luc introduced. The change itself is fine and
remain on RC5, it just make the bug trigger-able as side effect. The
real fix is submitted by Luc later.

> I ran the test case kill-unreachable-phi.c in my repository and did
> not see an infinite loop. My repository is way behind the latest
> changes in Sparse as I last synced up earlier this year so maybe I am
> not seeing this issue?

Try that on RC4 and you will see the dead loop.

Chris