ns-3.27 release update

classic Classic list List threaded Threaded
15 messages Options
Reply | Threaded
Open this post in threaded view
|

ns-3.27 release update

Tom Henderson-2
I have some time next week to work on a release candidate, so I am
proposing to put one out by next Wed. Sept 13, and make the release
shortly thereafter if the release candidate tests and documentation go
smoothly.

We haven't completely finished on the two issues/goals that I had for
this release (replacement of gccxml with CastXML for API scanning, and
zero Doxygen warnings) but we are close.  The API scanning isn't fully
automated yet but I think it is in a state where maintainers can
generate new bindings if needed.  We are closing in on our goal of zero
Doxygen warnings and probably can get most of the way there by next week.

Please let me know if you want to get a patch merged in the coming
week.  Stefano seems to be close to a final patch for the issue
described in this bug:
https://www.nsnam.org/bugzilla/show_bug.cgi?id=2751; we're likely to
merge something soon.  In the meantime, I will be focusing on issues in
the tracker in LAST CALL state, and any other bugs with patches ready to
go.

- Tom


Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Natale Patriciello
On 05/09/17 at 09:14pm, Tom Henderson wrote:

> I have some time next week to work on a release candidate, so I am proposing
> to put one out by next Wed. Sept 13, and make the release shortly thereafter
> if the release candidate tests and documentation go smoothly.
>
> We haven't completely finished on the two issues/goals that I had for this
> release (replacement of gccxml with CastXML for API scanning, and zero
> Doxygen warnings) but we are close.  The API scanning isn't fully automated
> yet but I think it is in a state where maintainers can generate new bindings
> if needed.  We are closing in on our goal of zero Doxygen warnings and
> probably can get most of the way there by next week.
>
> Please let me know if you want to get a patch merged in the coming week.
> Stefano seems to be close to a final patch for the issue described in this
> bug: https://www.nsnam.org/bugzilla/show_bug.cgi?id=2751; we're likely to
> merge something soon.  In the meantime, I will be focusing on issues in the
> tracker in LAST CALL state, and any other bugs with patches ready to go.
>

Hello Tom,

I wonder if some work of GSoC can be included or it is better to wait
the next release. Specifically, I would like to take some time on the
weekend to review from the beginning all the work done with ECN and
Prague at TCP level.

Nat
Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Tom Henderson-2

> Hello Tom,
>
> I wonder if some work of GSoC can be included or it is better to wait
> the next release. Specifically, I would like to take some time on the
> weekend to review from the beginning all the work done with ECN and
> Prague at TCP level
We should target the next release for the rest of our GSoC code; it is
still undergoing review and comment resolution.

- Tom
Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Mohit P. Tahiliani
In reply to this post by Tom Henderson-2
Hi Tom and Nat,

If possible, it would be great to merge the following in ns-3.27 release:

+ ECN implementation for TCP [1] (has been reviewed by Natale earlier, and
recently by Tom)

+ Tag to notify lower layer of TCP Pacing information [2] (needed to
proceed with the implementation of BBR)

+ Implementation of TCP Low Priority in ns-3 [3] (Natale has reviewed it a
few times, one final review might be required for test suite)

I will finish reviewing the following over the next two days:

+ ECN support in CoDel (patch by Stefano [4])

+ ECN support in PIE (patch by Shravya [5])

Let me know your opinion.

Thanks and Regards,
Mohit P. Tahiliani

[1] https://codereview.appspot.com/314790043/
[2] https://codereview.appspot.com/312560043/
[3] https://codereview.appspot.com/319240043/
[4] https://codereview.appspot.com/316180043/
[5] https://codereview.appspot.com/319210043/

On Thu, Sep 7, 2017 at 12:46 AM, Natale Patriciello <
[hidden email]> wrote:

> On 05/09/17 at 09:14pm, Tom Henderson wrote:
> > I have some time next week to work on a release candidate, so I am
> proposing
> > to put one out by next Wed. Sept 13, and make the release shortly
> thereafter
> > if the release candidate tests and documentation go smoothly.
> >
> > We haven't completely finished on the two issues/goals that I had for
> this
> > release (replacement of gccxml with CastXML for API scanning, and zero
> > Doxygen warnings) but we are close.  The API scanning isn't fully
> automated
> > yet but I think it is in a state where maintainers can generate new
> bindings
> > if needed.  We are closing in on our goal of zero Doxygen warnings and
> > probably can get most of the way there by next week.
> >
> > Please let me know if you want to get a patch merged in the coming week.
> > Stefano seems to be close to a final patch for the issue described in
> this
> > bug: https://www.nsnam.org/bugzilla/show_bug.cgi?id=2751; we're likely
> to
> > merge something soon.  In the meantime, I will be focusing on issues in
> the
> > tracker in LAST CALL state, and any other bugs with patches ready to go.
> >
>
> Hello Tom,
>
> I wonder if some work of GSoC can be included or it is better to wait
> the next release. Specifically, I would like to take some time on the
> weekend to review from the beginning all the work done with ECN and
> Prague at TCP level.
>
> Nat
>
Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Natale Patriciello
2017-09-06 21:56 GMT+02:00 Mohit P. Tahiliani <[hidden email]>:

> + Tag to notify lower layer of TCP Pacing information [2] (needed to proceed
> with the implementation of BBR)

I would not use a Tag, but instead, to implement BBR I would implement
the pacing directly in the TCP (as it is from kernel 4.13, recently
released). Then we can work out a way (which do not involve Tags) to
make the two layers communicate.

> + ECN implementation for TCP [1] (has been reviewed by Natale earlier, and recently by Tom)
> + Implementation of TCP Low Priority in ns-3 [3] (Natale has reviewed it a few times, one final review might be required for test suite)

I will take another look in the weekend for these two.


Nat
Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Natale Patriciello
In reply to this post by Tom Henderson-2
On 06/09/17 at 07:35pm, Tom Henderson wrote:
> We should target the next release for the rest of our GSoC code; it is still
> undergoing review and comment resolution.
>

Ok, so for what regards TCP:

(release blocking):

- #2781 (fixed just before this e-mail)
- #2769 misses the test. If Michele can work it out, better, otherwise
  Tuesday I will commit the fix without the patch (waiting for a later
  merge of the test).


(medium priority, in inverse order of fixing difficulty):

- #2652 (can be merged if no objections are raised)
- #2677 (it only needs whitespace fixes and a bit more documentation in
  my opinion)
- #2653 (ledbat authors or me, matter of minutes and 1 hour for the test
  case)
- #2678 (waiting response from authors)


ECN, GSoC work, and pacing will not be included in this release.
If I forgot something, please report it :)

Nat

Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Tom Henderson-2


On 9/10/17 6:36 AM, Natale Patriciello wrote:

> On 06/09/17 at 07:35pm, Tom Henderson wrote:
>> We should target the next release for the rest of our GSoC code; it is still
>> undergoing review and comment resolution.
>>
> Ok, so for what regards TCP:
>
> (release blocking):
>
> - #2781 (fixed just before this e-mail)
> - #2769 misses the test. If Michele can work it out, better, otherwise
>    Tuesday I will commit the fix without the patch (waiting for a later
>    merge of the test).

Agreed

>
>
> (medium priority, in inverse order of fixing difficulty):
>
> - #2652 (can be merged if no objections are raised)
> - #2677 (it only needs whitespace fixes and a bit more documentation in
>    my opinion)
> - #2653 (ledbat authors or me, matter of minutes and 1 hour for the test
>    case)
> - #2678 (waiting response from authors)

My sense is that #2678 will require time to address the review comments
(unanswered for some time).  I would prioritize 2653 in the above list
(bug fix) while 2652 and 2677 are new features.

I would also suggest we try to fix open bugs with patches if possible.  
Could you look at and try to merge these if ready?

#2133 TCP close, receive data in FIN_WAIT_2 state
#2716 (TCP socket ignoring FIN flag when in FIN_WAIT state)
#2285 (receive callback to fire before connection is established)

There are some that I'd like to fix if it can be patched this week (I
probably own the first three in the below list):

#2565 (do not wait 2*MSL to notify socket close)
#2649 (NewReno when SACK is disabled-- I probably have action item on
this issue)
#2750 (NotifyNormalClose() called upon error conditions)
#2278 (TCP doesn't close properly in some scenarios)

Tom
Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Tom Henderson-2
In reply to this post by Tom Henderson-2
On 09/05/2017 02:14 PM, Tom Henderson wrote:
> I have some time next week to work on a release candidate, so I am
> proposing to put one out by next Wed. Sept 13, and make the release
> shortly thereafter if the release candidate tests and documentation go
> smoothly.

I'm holding this for the following changes; hopefully done tomorrow:
- apply final patches for traffic control
(https://www.nsnam.org/bugzilla/show_bug.cgi?id=2751)
- some additional changes to support CastXML
- the TCP work mentioned earlier in the week

Tom
Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Natale Patriciello
In reply to this post by Tom Henderson-2
Hello,

I have done some work. Tom, please, take a look at the patches in the
bug listing and if you agree merge/commit them (I do not have access
to my machine until Friday night). Thanks!

2017-09-11 0:27 GMT+02:00 Tom Henderson <[hidden email]>:
>
>
> On 9/10/17 6:36 AM, Natale Patriciello wrote:
>>
>> - #2769 misses the test. If Michele can work it out, better, otherwise
>>    Tuesday I will commit the fix without the patch (waiting for a later
>>    merge of the test).

#2769 patch without the test can be merged.

>> - #2653 (ledbat authors or me, matter of minutes and 1 hour for the test
>>    case)

fix in the bugtracker without the test

> #2133 TCP close, receive data in FIN_WAIT_2 state

fix in the bugtracker

> #2716 (TCP socket ignoring FIN flag when in FIN_WAIT state)

The fix for #2133 should work in this case, but I'm not totally sure
because of missing test

> #2285 (receive callback to fire before connection is established)

Can be merged

> #2565 (do not wait 2*MSL to notify socket close)
This can be merged with a big note in the release notes ... it changes
a well-established behavior

> #2649 (NewReno when SACK is disabled-- I probably have action item on this
> issue)
> #2750 (NotifyNormalClose() called upon error conditions)
> #2278 (TCP doesn't close properly in some scenarios)
>>
>> - #2652 (can be merged if no objections are raised)
>> - #2677 (it only needs whitespace fixes and a bit more documentation in
>>    my opinion)
>> - #2678 (waiting response from authors)

I didn't have time to look into these. All the patches in the
bugtracker do not contain notes in the release notes, and I realized I
also didn't write anything for the bug #2781.

Thanks again!
Nat
Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Tom Henderson-2
In reply to this post by Tom Henderson-2
On 09/13/2017 10:39 PM, Tom Henderson wrote:
> On 09/05/2017 02:14 PM, Tom Henderson wrote:
>> I have some time next week to work on a release candidate, so I am
>> proposing to put one out by next Wed. Sept 13, and make the release
>> shortly thereafter if the release candidate tests and documentation
>> go smoothly.
>
> I'm holding this for the following changes; hopefully done tomorrow:
> - apply final patches for traffic control
> (https://www.nsnam.org/bugzilla/show_bug.cgi?id=2751)

We did apply the above but hit a snag in the MacOS build with it, so we
need to sort that out before proceeding.  I'll update again tomorrow.

- Tom

Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Tom Henderson-2
On 09/14/2017 10:47 PM, Tom Henderson wrote:

> On 09/13/2017 10:39 PM, Tom Henderson wrote:
>> On 09/05/2017 02:14 PM, Tom Henderson wrote:
>>> I have some time next week to work on a release candidate, so I am
>>> proposing to put one out by next Wed. Sept 13, and make the release
>>> shortly thereafter if the release candidate tests and documentation
>>> go smoothly.
>>
>> I'm holding this for the following changes; hopefully done tomorrow:
>> - apply final patches for traffic control
>> (https://www.nsnam.org/bugzilla/show_bug.cgi?id=2751)
>
> We did apply the above but hit a snag in the MacOS build with it, so
> we need to sort that out before proceeding.  I'll update again tomorrow.

I believe that all of the issues with the traffic control upgrade have
been resolved.  I am still working on one more TCP-related bug (2649)
before moving towards a release candidate.  There have been several wifi
issues posted and updated in the tracker in the past couple of days;
maybe we can try to address one or two of them as well.  I'll try to
provide another update in a day or so.

- Tom


Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Tom Henderson-2

> I believe that all of the issues with the traffic control upgrade have
> been resolved.  I am still working on one more TCP-related bug (2649)
> before moving towards a release candidate.  There have been several
> wifi issues posted and updated in the tracker in the past couple of
> days; maybe we can try to address one or two of them as well.  I'll
> try to provide another update in a day or so.

The above items have been finished, I'm now working on making a release
candidate based on the current ns-3-dev, once I finish updating all of
the Python bindings and test on our various buildslaves tomorrow.

- Tom

Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Stefano Avallone
Hi Tom,

may I push the attached commit? It makes a few simple changes to one of the
traffic-control examples (codel-vs-pfifo-basic-test) in order to set more
reasonable default (device queue size set to 10 instead of 1000, otherwise no
difference can be noticed between codel and pfifo) and allow to change the
queue disc capacity. The patch was prepared by Pasquale after a user notified
us about this issue.

Another question, out of curiosity. Gustavo's pybindgen repo does not include
the latest changes by Ankit (which I believe are necessary for castxml to
work). What is the plan? Will these changes eventually merged or are we going
to switch to Ankit's repo?

Thanks,
Stefano


On mercoledì 27 settembre 2017 06:39:24 CEST Tom Henderson wrote:

> > I believe that all of the issues with the traffic control upgrade have
> > been resolved.  I am still working on one more TCP-related bug (2649)
> > before moving towards a release candidate.  There have been several
> > wifi issues posted and updated in the tracker in the past couple of
> > days; maybe we can try to address one or two of them as well.  I'll
> > try to provide another update in a day or so.
>
> The above items have been finished, I'm now working on making a release
> candidate based on the current ns-3-dev, once I finish updating all of
> the Python bindings and test on our various buildslaves tomorrow.
>
> - Tom


0001-traffic-control-Set-sane-defaults-for-codel-vs-pfifo.patch (3K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Tom Henderson-2
On 09/27/2017 09:46 AM, Stefano Avallone wrote:
> Hi Tom,
>
> may I push the attached commit? It makes a few simple changes to one of the
> traffic-control examples (codel-vs-pfifo-basic-test) in order to set more
> reasonable default (device queue size set to 10 instead of 1000, otherwise no
> difference can be noticed between codel and pfifo) and allow to change the
> queue disc capacity. The patch was prepared by Pasquale after a user notified
> us about this issue.

Yes, please push it now.  I am waiting for the bindings testing to
finish before pushing those updates.

>
> Another question, out of curiosity. Gustavo's pybindgen repo does not include
> the latest changes by Ankit (which I believe are necessary for castxml to
> work). What is the plan? Will these changes eventually merged or are we going
> to switch to Ankit's repo?

I haven't checked in with Gustavo about it lately; if he is not ready to
accept the file (castxmlparser.py) yet, we can pull it from Ankit's repo.

I'm going to also push the 'ns3 patch to remove anonymous typedefs' in
the issue tracker for issue 2451.

- Tom

Reply | Threaded
Open this post in threaded view
|

Re: ns-3.27 release update

Tom Henderson-2
In reply to this post by Stefano Avallone
On 09/27/2017 09:46 AM, Stefano Avallone wrote:
> Hi Tom,
>
> may I push the attached commit? It makes a few simple changes to one of the
> traffic-control examples (codel-vs-pfifo-basic-test) in order to set more
> reasonable default (device queue size set to 10 instead of 1000, otherwise no
> difference can be noticed between codel and pfifo) and allow to change the
> queue disc capacity. The patch was prepared by Pasquale after a user notified
> us about this issue.

Stefano, I went ahead and committed that patch, and now plan to let the
buildslaves work on ns-3-dev overnight, and if everything passes, I'll
make the release candidate.

- Tom