Re: BIND 10 #322: Process name in Python

BIND 10 Development do-not-reply at isc.org
Mon Sep 27 11:12:37 UTC 2010


#322: Process name in Python
-------------------------------+--------------------------------------------
      Reporter:  fujiwara      |        Owner:  jinmei   
          Type:  enhancement   |       Status:  reviewing
      Priority:  major         |    Milestone:           
     Component:  Unclassified  |   Resolution:           
      Keywords:                |    Sensitive:  0        
Estimatedhours:  0.0           |        Hours:  0        
      Billable:  1             |   Totalhours:  3.0      
      Internal:  0             |  
-------------------------------+--------------------------------------------
Changes (by vorner):

  * owner:  vorner => jinmei


Comment:

 Replying to [comment:6 jinmei]:
 > Replying to [comment:5 vorner]:
 > > That will be a little bit of a problem. Many tools (top, pstree, …)
 cut too long names, so we would have many processes that would look like
 „/usr/share/local/libe“ and the important information which actual process
 it is would be not visible (and I think this is more common than having
 two binds).
 > >
 > Hmm, but from quick checks the top command I can use (including one
 available on bind10.isc.org) seems to cut the path itself.  I've never
 used pstree, but the one installed on bind10.isc.org seems to do the same
 thing.

 Well, there seem to be two (or more) process names or something, it cut
 the path if it is the name automatically assigned to it by system. But if
 I try to assign it by the setproctitle library, it does not cut the path,
 it cuts it from the end if it is too long, no matter if it is just some
 text or real path.

 > > On the other hand, which installation it is is valuable only for the
 boss. So maybe boss could have some more clever naming (like bind10
 <version>) with possibility to provide custom name by --prettyname, while
 other „slave“ processes would have just their basename? Would that be
 suitable?
 > >
 > Perhaps, and one possible compromise would be to add the path to bind10
 with --prettyname while keeping others short.

 Ok, BoB now has full path/arg 0, others have a basename and BoB has the
 --pretty-name parameter.

 > No it doesn't (or I was not clear in my previous comment and we are
 talking about different things).  For example, in bind10/Makefile.in we
 have
 > {{{
 > sbin_SCRIPTS = bind10
 > }}}
 >
 > So the process_test looks for {top_dir}/src/bin/bind10/bind10.
 >
 > But if we do ./configure then 'make check' immediately, "bind10" doesn't
 exist at the time the test is invoked because make first goes down to
 src/lib and then src/bin.  As a result you'd see:

 Ah, sorry, I misunderstood what you ment by vanilla tree. I moved it into
 src/bin/tests and made sure the scripts are created first. I didn't
 reallize anyone would like to run make check and not run make first.

 The make check fails now, if it is run directly, but with something
 unrelated that is fixed in trunk. If I merge them locally, it is fine (I
 didn't put the merge in SVN, since then the diff of a branch is harder to
 create).

 Thanks

-- 
Ticket URL: <https://bind10.isc.org/ticket/322#comment:7>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development


More information about the bind10-tickets mailing list