From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Tue, 13 May 2014 06:58:06 -0700 From: quanstro To: plan9port-dev@googlegroups.com Cc: quanstro@gmail.com, Fans of the OS Plan 9 from Bell Labs <9fans@9fans.net> Message-Id: <4e562e73-7f55-48f1-a388-1b73d84078ed@googlegroups.com> In-Reply-To: References: <849c2fe8-20f3-43d5-84c9-b50165958102@googlegroups.com> <0b731054-ea10-41e4-ab01-5f56ff6816f1@googlegroups.com> Subject: Re: acme, upas commits MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_Part_1856_27989063.1399989486713" Topicbox-Message-UUID: e6291892-ead8-11e9-9d60-3106f5b1d025 ------=_Part_1856_27989063.1399989486713 Content-Type: text/plain; charset=UTF-8 right. as i mentioned, the structure of acme had been to keep all the threads that might otherwise need locking in the same proc so as to use cooperative multitasking to avoid the need for explicit locking. therefore, i would not expect this bug to happen unless this code is either somehow exceptional, or acme was working in to procs when it should not be, or perhaps threads are not as cooperative as they might be. without resolution on this issue, it would seem likely that other bugs are lurking. even if they have not yet been found. that is, it might be that a more systemic fix may be necessary. - erik On Tuesday, May 13, 2014 9:34:17 AM UTC-4, Alexander Sychev wrote: > > Are there other similar bugs? > > Actually, I was digging only the amail's problem. > Now I can't recall all details (it was few months ago), but I had saw > parallel calls of regexp functions without any lock and I thought it should > be such way. > > > On Tue, May 13, 2014 at 4:44 PM, quanstro > > wrote: > >> did you track down why the unexpected concurrency was happening? this >> could result in other bugs as well. >> >> - erik >> >> >> On Tuesday, May 13, 2014 8:34:29 AM UTC-4, Alexander Sychev wrote: >> >>> There were problems in p9p under linux, when amail makes simultaneous >>> searches in different windows. >>> >>> >>> On Tue, May 13, 2014 at 4:23 PM, quanstro wrote: >>> >>>> in theory the threads that interact with regular expressions in acme >>>> are all cooperatively scheduled. have you seen a case where they >>>> are not? did pthreads blow up the model? >>>> >>>> - erik >>>> >>>> >>>> On Tuesday, May 13, 2014 7:11:02 AM UTC-4, Alexander Sychev wrote: >>>>> >>>>> Hi, >>>>> >>>>> I have commited some changes in my experimental fork of plan9port. >>>>> https://bitbucket.org/santucco/plan9port >>>>> >>>>> For acme I made thread safe processing of regular expressions >>>>> (actually by creating some context). >>>>> For upas/nfs I made a monitoring of all mailboxes and a decoding of >>>>> international names of mailboxes. >>>>> >>>>> I'm living with these changes for some time, they are working. >>>>> >>>>> P.S. >>>>> By the way, new log feature is excellent, it's a time to make more >>>>> applications for acme :-) >>>>> -- >>>>> Best regards, >>>>> santucco >>>>> >>>> -- >>>> >>>> --- >>>> You received this message because you are subscribed to the Google >>>> Groups "plan9port-dev" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to plan9port-de...@googlegroups.com. >>>> >>>> For more options, visit https://groups.google.com/d/optout. >>>> >>> >>> >>> >>> -- >>> Best regards, >>> santucco >>> >> -- >> >> --- >> You received this message because you are subscribed to the Google Groups >> "plan9port-dev" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to plan9port-de...@googlegroups.com . >> For more options, visit https://groups.google.com/d/optout. >> > > > > -- > Best regards, > santucco > -- --- You received this message because you are subscribed to the Google Groups "plan9port-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to plan9port-dev+unsubscribe@googlegroups.com. For more options, visit https://groups.google.com/d/optout. ------=_Part_1856_27989063.1399989486713 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
right.  as i mentioned, the structure of acme had bee= n to keep all the
threads that might otherwise need locking in the same= proc so as to
use cooperative multitasking to avoid the need for= explicit locking.
therefore, i would not expect this bug to happ= en unless this code is
either somehow exceptional, or acme was wo= rking in to procs when it
should not be, or perhaps threads are n= ot as cooperative as they might
be.

with= out resolution on this issue, it would seem likely that other bugs
are lurking.  even if they have not yet been found.  that is, i= t might be
that a more systemic fix may be necessary.
<= br>
- erik

On Tuesday, May 13, 2014 9:34= :17 AM UTC-4, Alexander Sychev wrote:
Are there other similar bugs?

Ac= tually, I was digging only the amail's problem. 
Now I can't= recall all details (it was few months ago), but I had saw parallel calls o= f regexp functions without any lock and I thought it should be such way.


On Tue, May 13, 2014 at 4:44 = PM, quanstro <qua= n...@gmail.com> wrote:
did you track down why the = unexpected concurrency was happening?  this
could result in other = bugs as well.

- erik


On Tuesday, May 13, 2014 8:34:29 AM = UTC-4, Alexander Sychev wrote:
There were problems in p9p under linux, when amail ma= kes  simultaneous searches in different windows. 


On Tue, May 13, 2014 at 4:23 PM, q= uanstro <quan...@gmail.com> wrote:
in theory the th= reads that interact with regular expressions in acme
are all cooperativ= ely scheduled.  have you seen a case where they
are not?  did pthreads blow up the model?

- erik


On Tuesday, May 13, 2014 7:11:02 AM UTC-4, Alexan= der Sychev wrote:
Hi,

I have commited some changes in my = experimental fork of plan9port.

For acme I made thread safe processing of regular expressions (a= ctually by creating some context).
For upas/nfs I made a monitori= ng of all mailboxes and a decoding of international names of mailboxes.

I'm living with these changes for some time, they are w= orking. 

P.S.
By the way, = new log feature is excellent, it's a time to make more applications for acm= e :-)
--
Best regards,
  santucco

--

---
You received this message because you are subscribed to the Google Groups "= plan9port-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an e= mail to plan9port-de...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.



--
Best regards,
  santucco

--

---
You received this message because you are subscribed to the Google Groups "= plan9port-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an e= mail to plan9port-de...@googlegro= ups.com.
For more options, visit https://groups.google.com/d/optout.



--
= Best regards,
  santucco

--

---
You received this message because you are subscribed to the Google Groups &= quot;plan9port-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an e= mail to plan9= port-dev+unsubscribe@googlegroups.com.
For more options, visit http= s://groups.google.com/d/optout.
------=_Part_1856_27989063.1399989486713--