Re: [Monetdb-developers] pathfinder release: bugfixing opportunity
Forgot to answer my question: The problem with the stack that was too small has been fixed. The StandOff-joins now increase the stack-size whenever it grows out-of-bound. The StandOff-axes in the stable branch still have two bugs to fix that I know of: - sorting of the final result (with multiple iters, the result is not sorted at the moment). The current work-around is adding a self-step directly after the StandOff-step. - David Ahn mentioned another bug, when nesting 3 for-loops, each using the select-narrow-axis the output is garbage, but I don't know whether this bug is related to the sorting, or perhaps to another part of MonetDB/XQuery. Still have to figure this out. Greetings, Wouter
-----Original Message----- From: Stefan Manegold [mailto:Stefan.Manegold@cwi.nl] Sent: Thursday, June 08, 2006 9:42 AM To: monetdb-developers@lists.sourceforge.net Cc: Wouter Alink (DT); 'Jan Flokstra'; 'Maurice Van Keulen' Subject: Re: pathfinder release: bugfixing opportunity
[Please use the MonetDB-developers mailing list (monetdb-developers@lists.sourceforge.net) for this kind of discussions, and make sure you're subscribed to it (with your current email address). Thanks!]
I suggest to not checkin any changes to the XQuery_0-10 stable branch of pathfinder, but to wait until we (Niels, Sjoerd, me) create the new XQuery_0-12 stable branch (together with the new MonetDB_4-12 & SQL_2-12 release branches) early next week. Of course, you're free and welcome to try to compile pathfinder's stable branch (XQuery_0-10) with the current MonetDB 4.11 development trunk to sort-out the problems and start fixing them off-line.
The short-term "feature-freeze" for the "intermediate release" applied only to MonetDB & sql; as Peter explained, for pathfinder we plan only porting of XQuery_0-10 to comply/compile with MonetDB 4.11 as well as applying bugfixes (including back-porting bug fixes from XQuery 0.11 as far as necessary/possible).
-----Original Message----- Hi,
I am not against including the compiler directory, but we must be careful with: - milprint-summer, I think it can't be rolled forward (all those tijah changes, and the update changes) - update syntax in the parser (it should not be recognized yet) - burkowski changes in the parser (may or may not be a
below)
You are probably right that removing code probably is as much work as finding the respective bug fixes. I just skipped through the commit mails of the last 4 months and looked up checkins that
On Thu, Jun 08, 2006 at 12:44:03AM +0200, Jan Rittinger wrote: problem, see mainly fix bugs.
So here's a (probably incomplete) list: milprint_summer.c 182->183, 185->186, 191->192, 208->209, 209->210, 210->211, 217->218, 219->220, 220->221 shredder.mx 77->78, 81->82 pf_support.mx 126->127 subtyping.c 33->34 fs.brg 33->34
the rule is that if you feel like it and it takes little time, it may be welcome.
I identified the following open bugs op the stable branch:
1447195 PF: crash server with XMark 1446426 fn:doc() call in nested element construction fails 1339844 PF: A parent step into the void 1459974 XQuery: parent-step into all 1491723 wrong position numbering in sequence after predicate
and:
1410662 PF: tests tests/XQuery/Tests/typing[12].xq fail 1365379 PF: "stable order by" 1351516 PF: attribute after non-attribute content 1207048 XQuery: crashtest
those that are "not (yet?) fixable" should be close with the proper comment/explanation.
Stefan
will try to compile pathfinder tomorrow and look into
these issues.
feel free to claim ony of these bugs..
I also identified the following closed bugs on the stable branch. They appeared to have been applied only to the HEAD, so I backported them to the Stable. This must still be checked:
1490847 problem with functions in xquery QUESTION TO JAN RITTINGER: was this really the small change mentioned in the bug report? I applied it.
it's in the above patch list
1454407 Mserver does not close XQuery sockets QUESTION TO NIELS NES: the patch in the bug report was ok? I applied it.
this one I don't know what to do with:
1463558 wrong result when passing nodelist to function QUESTION TO JAN RITTINGER: what was the bug-fix here?
it's in the above patch list
QUESTION TO WOUTER: I reported two problems with
we roll forward pf_burk as it is now to the stable, and get those fixes as well as the rename?
As the release date is near (I think Niels aims for less than two weeks), please treat this with urgency, or do not treat
select-narrow. Can this at all..
best,
Peter
Two bugs that might pop up if we roll forward the MonetDB
batcache) to Stable pathfinder:
1499969 PF: tests/XQuery/Tests/typeswitch[34].xq: property errors 1499962 PF: XMark Q10: "HASHremove: Assertion `0' failed."
-----Original Message----- From: Jan Rittinger [mailto:rittinge@in.tum.de] Sent: 07 June 2006 23:46 To: 'Peter Boncz'; wouter@holmes.nl; 'Torsten Grust'; 'Jens Teubner'; 'Jan Flokstra'; 'Maurice Van Keulen'; 'Wouter Alink' Cc: 'Sjoerd Mullender'; 'Stefan Manegold'; Niels.Nes@cwi.nl Subject: RE: pathfinder release: bugfixing opportunity
I have no problem postponing the pf release.
However I think that I (and also others) fixed quite a number of 'important' pathfinder bugs. In addition some small but very useful extensions sneaked into the code (e.g. we now support all kinds of typeswitches). It would be a pitty to ignore them in the coming release.
Is it perhaps possible to include the compiler directory (with small adjustments to align it with pf runtime and MonetDB), but leave out changes of the runtime?
Jan
-----Original Message----- From: Peter Boncz [mailto:boncz@cwi.nl] Sent: Wednesday, June 07, 2006 9:41 PM To: wouter@holmes.nl; Torsten Grust; Jens Teubner; Jan Rittinger; Jan Flokstra; Maurice Van Keulen; Wouter Alink Cc: Sjoerd Mullender; Stefan Manegold; niels@cwi.nl Subject: pathfinder release: bugfixing opportunity
Hi,
Today we had a MonetDB meeting about the release. We decided to have a release now, but without Pathfinder.
For the Pathfinder release with updates, multi-documents and soap the realistic time-frame becomes *september* (well, maybe maybe august), but there will be a feature freeze around SIGMOD (june 25). We anticipate a long testing and debugging stage. Sorry, but this is reality. No SIGMOD release.
As MonetDB/SQL and the alpha Monet5 do not want to wait any longer just to roll through the accumulated new features and fixes, there is also an imminent release. As we release always all components to keep compatability and the test web easily manageble in a single Stable an single Current branch, the new release will include the *stable* branch of pathfinder.
There will be a (very) small porting effort to make stable-pathfinder work on head-monet. However, one of the tasks ahead for a pathfinder release anyway is bug-fixing. Thus, we can take this opportunity to identify quick-win bugs in pathfinder and release those. We can also look at the bugs fixed in pathfinder in the past period, and ask ourselfves whether rollling them through is useful and little work.
I guess the freeze for the direct release is only within days, and I recommend you all not to spend much time on it. At least the new pathfinder will contain the BATuselect fix in the serialization (admittedly not as fast as Jan R.'s current implementation but still better than before).
As the experiences with it have been positive, I propose
to build pathfinder with burkowski enabled. Thus we can
head (with this time point the
guy who wants windows with burkowski (and other like him) to the new binary release. Note that the functionality is by default still disabled; you need to set a configuration switch to yes to turn it on.
Peter
-- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | CWI, P.O.Box | 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |
participants (1)
-
Wouter Alink (DT)