This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH WIP 0/4] GDB enhancement on ITSET


On 09/13/2012 01:53 AM, Tom Tromey wrote:
I appreciate the work you did writing up this series and posting it.
Publishing a git repository would make sharing quite easy though...


That is a good idea, and I am setting it up on github. The tree will be ready in two or three days.


I read through this series and it all looked pretty reasonable to me.

Thanks.


Yao>   [PATCH 0/3] Basic ITSET
Yao>http://sourceware.org/ml/gdb-patches/2012-05/msg01079.html

I guess this is still blocked on some other work?

From the patch/diff's point of view, the answer is no. Patches 'Basic ITSET' should be applied cleanly on mainline, as they only add new files, and 'Basic ITSET' is only about creating/deleting itset. This part is quite isolated, and is not blocked by other work.


In order to go to the final destination of ITSET, we still have some steps in my mind to go, (Pedro may/must know more)

1) patches on 'run all-stop mode on top of non-stop target'.
Current patches are broken on 'software-single-step' ports, and we may need a per-thread software-single-step breakpoints (IIRC).
2) patches to replace 'current thread' and 'current inferior' with
'current itset'.
3) Finalize itset spec/syntax.


From this point of view, 'ITSET enhancement' and 'Basic ITSET' is blocked by these three steps.

I am thinking how to parallelize these steps, because looks the first two steps are complicated, and the latter steps are relatively easier.
Can we define a minimum set of ITSET spec/syntax first, in which we are sure that the spec/syntax will be unchanged? For example, the minimum spec can be only for cores, like c1, c2, and c3. Then, 'Basic ITSET' and 'ITSET enhancement' patches can go in. In parallel, we can continue three steps above. At the same time, we (or I) can start to investigate using ITSET in target side (in gdbserver).


In short, we can start from a minimum set of ITSET sepc and don't associate it with execution control, so that the starting point is relatively easy, and we can extend ITSET work gradually. Thoughts?

--
Yao


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]