You are viewing a plain text version of this content. The canonical link for it is here.
Posted to derby-dev@db.apache.org by Myrna van Lunteren <m....@gmail.com> on 2007/07/07 02:51:28 UTC
10.3 - to release or not to release?
Hi,
I've been looking over the fixed and open bugs for the 10.3 release,
and there is a lot of good stuff...
However, there are also some worrying aspects:
- There are a number of Critical bugs open, most without any
indication of getting a fix. See:
https://issues.apache.org/jira/secure/IssueNavigator.jspa?reset=true&pid=10594&priority=2&resolution=-1
- There are a number of intermittent test regressions so we're not
getting consistently clean test runs. Just look at the platform
testing list to see this:
http://wiki.apache.org/db-derby/TenThreePlatformTesting
In addition, there are 7 items which never received any Buddy testing...
Further, we still ran into a deterioration in the suites.All at one
point, requiring running with extra memory on many systems, which was
never addressed.
- There are a largish number of open bugs, many of which are marked
affects 10.3., i.e. affect the new functionality added with 10.3.
On the other hand, it seems to me some of the current functionality,
is eagerly awaited by the user community. A number of the critical
bugs are not new with 10.3., or are related to security, which may, or
may not affect a number of our users...And none of them have an
owner...
So, I intend to continue with my stated plan of generating a Release
Candidate and call a vote next Monday, but I would suggest we plan for
a bug fix release off the 10.3 branch to specifically drive down the
number of bugs...
Thx,
Myrna
Re: 10.3 - to release or not to release?
Posted by Rick Hillegas <Ri...@Sun.COM>.
Stanley Bradbury wrote:
> Myrna van Lunteren wrote:
>> Hi,
>>
>> I've been looking over the fixed and open bugs for the 10.3 release,
>> and there is a lot of good stuff...
>>
>> However, there are also some worrying aspects:
>> - There are a number of Critical bugs open, most without any
>> indication of getting a fix. See:
>> https://issues.apache.org/jira/secure/IssueNavigator.jspa?reset=true&pid=10594&priority=2&resolution=-1
>>
>>
>> - There are a number of intermittent test regressions so we're not
>> getting consistently clean test runs. Just look at the platform
>> testing list to see this:
>> http://wiki.apache.org/db-derby/TenThreePlatformTesting
>> In addition, there are 7 items which never received any Buddy testing...
>> Further, we still ran into a deterioration in the suites.All at one
>> point, requiring running with extra memory on many systems, which was
>> never addressed.
>>
>> - There are a largish number of open bugs, many of which are marked
>> affects 10.3., i.e. affect the new functionality added with 10.3.
>>
>> On the other hand, it seems to me some of the current functionality,
>> is eagerly awaited by the user community. A number of the critical
>> bugs are not new with 10.3., or are related to security, which may, or
>> may not affect a number of our users...And none of them have an
>> owner...
>>
>> So, I intend to continue with my stated plan of generating a Release
>> Candidate and call a vote next Monday, but I would suggest we plan for
>> a bug fix release off the 10.3 branch to specifically drive down the
>> number of bugs...
>>
>> Thx,
>> Myrna
>>
> Hi Myrna -
> Thanks for the through release report. I took a look at the critical
> bug list and am concerned about the regression that causes locking
> problems: DERBY-2892
> <https://issues.apache.org/jira/browse/DERBY-2892> - this is a
> regression of the DERBY-255 fix and, though it was introduced in 10.2,
> should be addressed quickly. Has anyone look at the checkin that
> might have caused the regression?
> DERBY-326 <https://issues.apache.org/jira/browse/DERBY-326> (svn 405037).
If someone volunteers to scratch this itch imminently, then I think we
should consider delaying the release. Otherwise, I would not let this
10.2 bug gate the release of other quality improvements. Hopefully this
issue will find a champion soon or at least in time for the 10.3
maintenance release.
Regards,
-Rick
Re: 10.3 - to release or not to release?
Posted by Myrna van Lunteren <m....@gmail.com>.
Regarding the export/import issue (DERBY-2436 / 2437) & possibility of
holding up the release for it I have to bring this up: we could choose
to switch off the blob/clob import/export functionality. That still
leaves the security hole.
I tend to think that the added benefit of the having blob/clob
import/export functionality does outweigh the increase in impact of
the pre-existing security hole in import/export.
Myrna
Re: 10.3 - to release or not to release?
Posted by Kathey Marsden <km...@sbcglobal.net>.
Daniel John Debrunner wrote:
> The Apache Way includes:
>
> - consistently high quality software
> - security as a mandatory feature
>
What does "mandatory" mean? Clearly network server was released without
being secure and IMHO is still not secure, so does "mandatory" mean we
should have never released it? I have long worried about what a world
wide probe on port 1527 might glean. Incremental security fixes have
meant and will continue to mean incompatibilities. One user said to me,
it is not *if* something will break when we upgrade, it is *what* will
break when we upgrade. So, is Network Server release quality and
security according to the Apache Way and if not, what should we do
about it?
Kathey
Re: 10.3 - to release or not to release?
Posted by Daniel John Debrunner <dj...@apache.org>.
Myrna van Lunteren wrote:
> One of the aspects that I let weigh in on deciding to call a vote or
> to hold it up for these critical issues, is that no one *has* signed
> up. I can theorize based on past track records & hope that certain
> people might step up to the plate, but as long as there's no firm
> assignment coupled with an indication from the responsible individual
> that it's going to be done within a reasonable time frame, I don't
> think it makes sense to hold up a release...
One other aspect is how does a release reflect on the Apache brand and
the Derby community. The Apache Way includes:
- consistently high quality software
- security as a mandatory feature
I would encourage folks to consider all the information about the
release before voting, not just a run of the functional tests on some
platforms. That's information from comments like Bryan's (quality is
getting better), the outstanding bugs, non-functional testing reports
from others, etc.
I'm slightly worried about the value given to the functional tests
passing, because in some cases we find a bug, make the test pass by
avoiding the bug in the test and enter a Jira bug report. Thus while we
(I hope) wouldn't make a release if the functional tests are failing, it
seems we are quite happy to mask those failures by entering bugs. I
understand why it's done (to avoid others wasting time to see if they
caused those failures, ie to have clean runs), but it does devalue the
functional tests a little. Or at least one must look at them along with
the outstanding bugs.
Dan.
Re: 10.3 - to release or not to release?
Posted by Oystein Grovlen - Sun Norway <Oy...@Sun.COM>.
Kathey Marsden wrote:
> Myrna van Lunteren wrote:
>> On 7/9/07, Stanley Bradbury <St...@gmail.com> wrote:
>>> >
>>> Hi Myrna -
>>> Thanks for the through release report. I took a look at the critical
>>> bug list and am concerned about the regression that causes locking
>>> problems:
>>> DERBY-2892 <https://issues.apache.org/jira/browse/DERBY-2892> - this is
>>> a regression of the DERBY-255 fix and, though it was introduced in 10.2,
>>> should be addressed quickly. Has anyone look at the checkin that might
>>> have caused the regression?
>>
>> No one has signed up.
> Hi Myrna,
> I personally tend to think the release should be held up for this fix.
> If we take the philosophy that we don't have to fix a regression because
> it regressed a while ago we will constantly lose ground.
> It looks like the fix though will be different in 10.2 versus 10.3. I
> am looking at 10.2, but am not totally confident that I can get it fixed
> in short order. I would most appreciate if someone familiar with lob
> locators could take a look at the 10.3 fix.
I will take a look at how this can be solved for 10.3. However, I do
not think this issue is severe enough to stop the release. We are
talking about a regression in 10.2 that was not discovered until over
half a year after it was released.
I am not saying that we do not need to fix this regression. I am saying
that 10.3.1 can be released without a fix, but that it should be fixed
by a subsequent 10.3 update release.
In my opinion, we have gotten so far in the release process now that
unless it is of significantly less quality than previous releases we
should ship the release.
--
Øystein
Re: 10.3 - to release or not to release?
Posted by Kathey Marsden <km...@sbcglobal.net>.
Myrna van Lunteren wrote:
> On 7/9/07, Stanley Bradbury <St...@gmail.com> wrote:
>> >
>> Hi Myrna -
>> Thanks for the through release report. I took a look at the critical
>> bug list and am concerned about the regression that causes locking
>> problems:
>> DERBY-2892 <https://issues.apache.org/jira/browse/DERBY-2892> - this is
>> a regression of the DERBY-255 fix and, though it was introduced in 10.2,
>> should be addressed quickly. Has anyone look at the checkin that might
>> have caused the regression?
>
> No one has signed up.
Hi Myrna,
I personally tend to think the release should be held up for this fix.
If we take the philosophy that we don't have to fix a regression because
it regressed a while ago we will constantly lose ground.
It looks like the fix though will be different in 10.2 versus 10.3. I
am looking at 10.2, but am not totally confident that I can get it fixed
in short order. I would most appreciate if someone familiar with lob
locators could take a look at the 10.3 fix.
Kathey
Re: 10.3 - to release or not to release?
Posted by Daniel John Debrunner <dj...@apache.org>.
Stanley Bradbury wrote:
> Myrna van Lunteren wrote:
>> On 7/9/07, Stanley Bradbury <St...@gmail.com> wrote:
>>> >
>>> Hi Myrna -
>>> Thanks for the through release report. I took a look at the critical
>>> bug list
I'm concerned about: DERBY-2893
INSERT and UPDATES succeed when permission has not been granted.
https://issues.apache.org/jira/browse/DERBY-2893
Security problem and possible regression from 10.2 (though could be a
test conversion issue).
Dan.
Re: 10.3 - to release or not to release?
Posted by Stanley Bradbury <St...@gmail.com>.
Myrna van Lunteren wrote:
> On 7/9/07, Stanley Bradbury <St...@gmail.com> wrote:
>> >
>> Hi Myrna -
>> Thanks for the through release report. I took a look at the critical
>> bug list and am concerned about the regression that causes locking
>> problems:
>> DERBY-2892 <https://issues.apache.org/jira/browse/DERBY-2892> - this is
>> a regression of the DERBY-255 fix and, though it was introduced in 10.2,
>> should be addressed quickly. Has anyone look at the checkin that might
>> have caused the regression?
>
> No one has signed up.
>
>> DERBY-326 <https://issues.apache.org/jira/browse/DERBY-326> (svn
>> 405037).
>>
>> This also concerns me because I assume the problem is because of the
>> collation differences between the SYS schema and the User Database - I
>> am not upto date on this issue - can it be resolved?
>
> You must be meaning a different bug? Sorry, but I don't understand.
> I've been trying to figure out which bug you *could* be referring to,
> but maybe you can just tell me.
>
>> DERBY-2896 <https://issues.apache.org/jira/browse/DERBY-2896> - It
>> sounds like DatabaseMetaData calls will fail when TERRITORY_BASED
>> collation is used. This will mean that many IDEs will not work with
>> these databases. Since this is a new feature it will not break existing
>> systems but will slow use of the feature.
>>
>
> Again, no one signed up.
>
> One of the aspects that I let weigh in on deciding to call a vote or
> to hold it up for these critical issues, is that no one *has* signed
> up. I can theorize based on past track records & hope that certain
> people might step up to the plate, but as long as there's no firm
> assignment coupled with an indication from the responsible individual
> that it's going to be done within a reasonable time frame, I don't
> think it makes sense to hold up a release...
>
> Myrna
>
Hi Myrna -
Bad formatting on that last message.
The reference to Derby-326 relates to DERBY-2892 (the first issue). A
comment in DERBY-2832 suggests that the changes made for DERBY-326
caused the regression reported as DERBY-2832. I was hoping that someone
knowledgeable about the performance improvements could quickly address
this issue.
My concern about comparisons between the SYS schema tables and the User
schemas related to the second issue of DatabaseMetaData failures.
I don't think there is anything for you to do at this point. The
community needs to look at these and determine if they will be highly
detrimental to the Derby release. I have no doubt that, if released
with these problems, 10.3 users will encounter these issues sooner
rather than later.
After some testing I may raise these concerns in a separate thread.
And thanks again for your very through risk-analysis that highlighted
these (and other) issues. It shows you are very much 'on top' of the
Release.
Re: 10.3 - to release or not to release?
Posted by Myrna van Lunteren <m....@gmail.com>.
On 7/9/07, Stanley Bradbury <St...@gmail.com> wrote:
> >
> Hi Myrna -
> Thanks for the through release report. I took a look at the critical
> bug list and am concerned about the regression that causes locking
> problems:
> DERBY-2892 <https://issues.apache.org/jira/browse/DERBY-2892> - this is
> a regression of the DERBY-255 fix and, though it was introduced in 10.2,
> should be addressed quickly. Has anyone look at the checkin that might
> have caused the regression?
No one has signed up.
> DERBY-326 <https://issues.apache.org/jira/browse/DERBY-326> (svn 405037).
>
> This also concerns me because I assume the problem is because of the
> collation differences between the SYS schema and the User Database - I
> am not upto date on this issue - can it be resolved?
You must be meaning a different bug? Sorry, but I don't understand.
I've been trying to figure out which bug you *could* be referring to,
but maybe you can just tell me.
> DERBY-2896 <https://issues.apache.org/jira/browse/DERBY-2896> - It
> sounds like DatabaseMetaData calls will fail when TERRITORY_BASED
> collation is used. This will mean that many IDEs will not work with
> these databases. Since this is a new feature it will not break existing
> systems but will slow use of the feature.
>
Again, no one signed up.
One of the aspects that I let weigh in on deciding to call a vote or
to hold it up for these critical issues, is that no one *has* signed
up. I can theorize based on past track records & hope that certain
people might step up to the plate, but as long as there's no firm
assignment coupled with an indication from the responsible individual
that it's going to be done within a reasonable time frame, I don't
think it makes sense to hold up a release...
Myrna
Re: 10.3 - to release or not to release?
Posted by Stanley Bradbury <St...@gmail.com>.
Myrna van Lunteren wrote:
> Hi,
>
> I've been looking over the fixed and open bugs for the 10.3 release,
> and there is a lot of good stuff...
>
> However, there are also some worrying aspects:
> - There are a number of Critical bugs open, most without any
> indication of getting a fix. See:
> https://issues.apache.org/jira/secure/IssueNavigator.jspa?reset=true&pid=10594&priority=2&resolution=-1
>
>
> - There are a number of intermittent test regressions so we're not
> getting consistently clean test runs. Just look at the platform
> testing list to see this:
> http://wiki.apache.org/db-derby/TenThreePlatformTesting
> In addition, there are 7 items which never received any Buddy testing...
> Further, we still ran into a deterioration in the suites.All at one
> point, requiring running with extra memory on many systems, which was
> never addressed.
>
> - There are a largish number of open bugs, many of which are marked
> affects 10.3., i.e. affect the new functionality added with 10.3.
>
> On the other hand, it seems to me some of the current functionality,
> is eagerly awaited by the user community. A number of the critical
> bugs are not new with 10.3., or are related to security, which may, or
> may not affect a number of our users...And none of them have an
> owner...
>
> So, I intend to continue with my stated plan of generating a Release
> Candidate and call a vote next Monday, but I would suggest we plan for
> a bug fix release off the 10.3 branch to specifically drive down the
> number of bugs...
>
> Thx,
> Myrna
>
Hi Myrna -
Thanks for the through release report. I took a look at the critical
bug list and am concerned about the regression that causes locking
problems:
DERBY-2892 <https://issues.apache.org/jira/browse/DERBY-2892> - this is
a regression of the DERBY-255 fix and, though it was introduced in 10.2,
should be addressed quickly. Has anyone look at the checkin that might
have caused the regression?
DERBY-326 <https://issues.apache.org/jira/browse/DERBY-326> (svn 405037).
This also concerns me because I assume the problem is because of the
collation differences between the SYS schema and the User Database - I
am not upto date on this issue - can it be resolved?
DERBY-2896 <https://issues.apache.org/jira/browse/DERBY-2896> - It
sounds like DatabaseMetaData calls will fail when TERRITORY_BASED
collation is used. This will mean that many IDEs will not work with
these databases. Since this is a new feature it will not break existing
systems but will slow use of the feature.
Re: 10.3 - to release or not to release?
Posted by Daniel John Debrunner <dj...@apache.org>.
Myrna van Lunteren wrote:
> Hi,
>
> I've been looking over the fixed and open bugs for the 10.3 release,
> and there is a lot of good stuff...
>
> However, there are also some worrying aspects:
FYI - some numbers:
Derby Open Code Bugs
2007/01/09 - 209
2007/07/06 - 278
Of those 278
89 marked affects 10.3/10.4 (at least, may affect early releases)
Of those 89
14 marked regression (ie. regression from an earlier release)
26 marked as Regression Test Failure (ie. non-clean regression test
issue)
6 critical
Make of it what you will,
Dan.
Re: 10.3 - to release or not to release?
Posted by Rick Hillegas <Ri...@Sun.COM>.
Bryan Pendleton wrote:
>> So, I intend to continue with my stated plan of generating a Release
>> Candidate and call a vote next Monday, but I would suggest we plan for
>> a bug fix release off the 10.3 branch to specifically drive down the
>> number of bugs...
>
> +1 to both ideas.
>
> From my own perspective, the increased number of bugs in the database
> reflects the increased testing and exposure that Derby has received
> over the last year, and is not an indication of a problem, rather the
> opposite; it has been my experience that the quality of Derby
> has improved substantially during this time and I think this can be
> hard to measure, but it is a very real result.
>
> Thanks very much for all your great work on the 10.3 release Myrna!
> It is greatly appreciated by us in the community.
>
> thanks,
>
> bryan
>
+1
I agree with everything Bryan says here. I would also like to thank
Myrna for all the hard work.
Regards,
-Rick
Re: 10.3 - to release or not to release?
Posted by Bryan Pendleton <bp...@amberpoint.com>.
> So, I intend to continue with my stated plan of generating a Release
> Candidate and call a vote next Monday, but I would suggest we plan for
> a bug fix release off the 10.3 branch to specifically drive down the
> number of bugs...
+1 to both ideas.
From my own perspective, the increased number of bugs in the database
reflects the increased testing and exposure that Derby has received
over the last year, and is not an indication of a problem, rather the
opposite; it has been my experience that the quality of Derby
has improved substantially during this time and I think this can be
hard to measure, but it is a very real result.
Thanks very much for all your great work on the 10.3 release Myrna!
It is greatly appreciated by us in the community.
thanks,
bryan
Re: 10.3 - to release or not to release?
Posted by Daniel John Debrunner <dj...@apache.org>.
Myrna van Lunteren wrote:
> Hi,
>
> I've been looking over the fixed and open bugs for the 10.3 release,
> and there is a lot of good stuff...
[snip]
> Further, we still ran into a deterioration in the suites.All at one
> point, requiring running with extra memory on many systems, which was
> never addressed.
I believe this one was due to new tests (trigger based) being added that
exposed an existing bug, not due to a change in the engine.
> - There are a largish number of open bugs, many of which are marked
> affects 10.3., i.e. affect the new functionality added with 10.3.
That's not the meaning of affects 10.3. All that means is the bug can be
seen in 10.3, in existing or new functionality. Might be a regression,
might not be.
[snip]
Dan.