Re: Improving OSGi/bnd build and much more...

[Mark Thomas]

On 10/06/2020 14:36, Raymond Auge wrote:
> Thanks for the feedback. I'll look at each point carefully.
> 
> I can report though that tests are looking relatively good. Only a few
> failures:
> 
> TEST-org.apache.tomcat.util.net.openssl.ciphers.TestCipher.APR.txt:Tests
> run: 3, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.057 sec
> TEST-org.apache.tomcat.util.net.openssl.ciphers.TestCipher.NIO2.txt:Tests run:
> 3, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.054 sec
> TEST-org.apache.tomcat.util.net.openssl.ciphers.TestCipher.NIO.txt:Tests
> run: 3, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.057 sec
> TEST-org.apache.tomcat.util.net.openssl.ciphers.TestOpenSSLCipherConfigurationParser.APR.txt:Tests
> run: 86, Failures: 10, Errors: 0, Skipped: 1, Time elapsed: 0.283 sec
> TEST-org.apache.tomcat.util.net.openssl.ciphers.TestOpenSSLCipherConfigurationParser.NIO2.txt:Tests
> run: 86, Failures: 10, Errors: 0, Skipped: 1, Time elapsed: 0.287 sec
> TEST-org.apache.tomcat.util.net.openssl.ciphers.TestOpenSSLCipherConfigurationParser.NIO.txt:Tests
> run: 86, Failures: 10, Errors: 0, Skipped: 1, Time elapsed: 0.281 sec

Probably related to the OpenSSL version in place. The tests are
sensitive to differences in OpenSSL provided by various distributions.

> TEST-org.apache.tomcat.util.res.TestStringManager.APR.txt:Tests run: 5,
> Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.013 sec
> TEST-org.apache.tomcat.util.res.TestStringManager.NIO2.txt:Tests run: 5,
> Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.012 sec
> TEST-org.apache.tomcat.util.res.TestStringManager.NIO.txt:Tests run: 5,
> Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.012 sec

Those look worthy of investigation.

Mark


> 
> :)
> 
> On Tue, Jun 9, 2020 at 2:54 PM Mark Thomas  > wrote:
> 
> On 09/06/2020 18:37, Raymond Auge wrote:
> > Hey All (particularly Markt whom I spoke with about this months and
> > months ago),
> >
> > Noting that bnd was already added to the build I took a crack at
> making
> > better use of it.
> >
> > Here are a few items and some questions in no particular order:
> >
> > - you don't use bndlib, can we remove that? bnd is fully
> standalone for
> > your purposes and is all you need
> 
> If we don't need it then +1 to dropping it.
> 
> > - bootstrap.jar in master seems to be missing the class
> > `org.apache.tomcat.util.ExceptionUtils` at least from a dependency
> > analysis point of view, should it be included in the files list?
> 
> No sure. That needs some more careful thought / analysis.
> 
> > - tomcat-embed-core.jar appears to be missing tribes dependency (at
> > least it refers to it but does not include it) Shouldn't it really
> > contain all of catalina (minus catalina-ant?)
> 
> That is deliberate.
> 
> > - files.tomcat-embed-core seems to reference an old package
> > org/apache/jk/**, can we remove that?
> 
> Yes, that can be removed.
> 
> > - I've split out the i18n LocalStrings_* (the locale specific
> ones) into
> > a separate output dir so that you wouldn't have to then filter
> them out,
> > it makes building their zips also cleaner, does that sound logical?
> 
> I'd want to see the patch for that first.
> 
> > - I moved a slew of concerns that bnd could handle out of ant and
> > directly into the bnd configurations, manifest filtering, etc.,
> bnd has
> > filtering capability built in and knows about all ant properties
> so this
> > eliminates ant work, comments on that?
> 
> Again, I'd want to see the patch for that.
> 
> > - generated all the Name sections for spec API packages instead of
> > maintaining those by hand (now they are complete)
> 
> Sounds good.
> 
> > - add all the correct OSGi metadata of course to the spec jars and to
> > everyone else.
> 
> Sounds good. Note the naming convention may have changed in the move
> from Java EE to Jakarta EE.
> 
> > - I've added bnd's JPMS build support to all the modules, still builds
> > on Java 8, and you get JPMS module-info almost for free (very small
> > amount of configurable bits)
> 
> Generally sounds good. My only concern is consistency for the spec JARs.
> 
> > Could someone provide the build instruction I'd need to do the most
> > throughout test build? I see that there's a .travis script which does
> > `ant -q test 2>&1 > ant-test.log` is that really all?
> 
> Yes, assuming you have properties set to point to a valid Tomcat Native
> build.
> 
> > Shall I create an SVN branch to explore this? Or would a git PR be
> enough?
> 
> I'd recommend tacking this one issue at a time. PRs would be fine for
> that. Some should be pretty quick.
> 
> Mark
> 
> -
> To unsubscribe, e-mail: 

Re: Improving OSGi/bnd build and much more...

[Raymond Auge]

Thanks for the feedback. I'll look at each point carefully.

I can report though that tests are looking relatively good. Only a few
failures:

TEST-org.apache.tomcat.util.net.openssl.ciphers.TestCipher.APR.txt:Tests
run: 3, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.057 sec
TEST-org.apache.tomcat.util.net.openssl.ciphers.TestCipher.NIO2.txt:Tests
run: 3, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.054 sec
TEST-org.apache.tomcat.util.net.openssl.ciphers.TestCipher.NIO.txt:Tests
run: 3, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.057 sec
TEST-org.apache.tomcat.util.net.openssl.ciphers.TestOpenSSLCipherConfigurationParser.APR.txt:Tests
run: 86, Failures: 10, Errors: 0, Skipped: 1, Time elapsed: 0.283 sec
TEST-org.apache.tomcat.util.net.openssl.ciphers.TestOpenSSLCipherConfigurationParser.NIO2.txt:Tests
run: 86, Failures: 10, Errors: 0, Skipped: 1, Time elapsed: 0.287 sec
TEST-org.apache.tomcat.util.net.openssl.ciphers.TestOpenSSLCipherConfigurationParser.NIO.txt:Tests
run: 86, Failures: 10, Errors: 0, Skipped: 1, Time elapsed: 0.281 sec
TEST-org.apache.tomcat.util.res.TestStringManager.APR.txt:Tests run: 5,
Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.013 sec
TEST-org.apache.tomcat.util.res.TestStringManager.NIO2.txt:Tests run: 5,
Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.012 sec
TEST-org.apache.tomcat.util.res.TestStringManager.NIO.txt:Tests run: 5,
Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0.012 sec

:)

On Tue, Jun 9, 2020 at 2:54 PM Mark Thomas  wrote:

> On 09/06/2020 18:37, Raymond Auge wrote:
> > Hey All (particularly Markt whom I spoke with about this months and
> > months ago),
> >
> > Noting that bnd was already added to the build I took a crack at making
> > better use of it.
> >
> > Here are a few items and some questions in no particular order:
> >
> > - you don't use bndlib, can we remove that? bnd is fully standalone for
> > your purposes and is all you need
>
> If we don't need it then +1 to dropping it.
>
> > - bootstrap.jar in master seems to be missing the class
> > `org.apache.tomcat.util.ExceptionUtils` at least from a dependency
> > analysis point of view, should it be included in the files list?
>
> No sure. That needs some more careful thought / analysis.
>
> > - tomcat-embed-core.jar appears to be missing tribes dependency (at
> > least it refers to it but does not include it) Shouldn't it really
> > contain all of catalina (minus catalina-ant?)
>
> That is deliberate.
>
> > - files.tomcat-embed-core seems to reference an old package
> > org/apache/jk/**, can we remove that?
>
> Yes, that can be removed.
>
> > - I've split out the i18n LocalStrings_* (the locale specific ones) into
> > a separate output dir so that you wouldn't have to then filter them out,
> > it makes building their zips also cleaner, does that sound logical?
>
> I'd want to see the patch for that first.
>
> > - I moved a slew of concerns that bnd could handle out of ant and
> > directly into the bnd configurations, manifest filtering, etc., bnd has
> > filtering capability built in and knows about all ant properties so this
> > eliminates ant work, comments on that?
>
> Again, I'd want to see the patch for that.
>
> > - generated all the Name sections for spec API packages instead of
> > maintaining those by hand (now they are complete)
>
> Sounds good.
>
> > - add all the correct OSGi metadata of course to the spec jars and to
> > everyone else.
>
> Sounds good. Note the naming convention may have changed in the move
> from Java EE to Jakarta EE.
>
> > - I've added bnd's JPMS build support to all the modules, still builds
> > on Java 8, and you get JPMS module-info almost for free (very small
> > amount of configurable bits)
>
> Generally sounds good. My only concern is consistency for the spec JARs.
>
> > Could someone provide the build instruction I'd need to do the most
> > throughout test build? I see that there's a .travis script which does
> > `ant -q test 2>&1 > ant-test.log` is that really all?
>
> Yes, assuming you have properties set to point to a valid Tomcat Native
> build.
>
> > Shall I create an SVN branch to explore this? Or would a git PR be
> enough?
>
> I'd recommend tacking this one issue at a time. PRs would be fine for
> that. Some should be pretty quick.
>
> Mark
>
> -
> To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
> For additional commands, e-mail: dev-h...@tomcat.apache.org
>
>

-- 
*Raymond Augé* 
 (@rotty3000)
Senior Software Architect *Liferay, Inc.* 
 (@Liferay)

Re: Improving OSGi/bnd build and much more...

[Mark Thomas]

On 09/06/2020 18:37, Raymond Auge wrote:
> Hey All (particularly Markt whom I spoke with about this months and
> months ago),
> 
> Noting that bnd was already added to the build I took a crack at making
> better use of it.
> 
> Here are a few items and some questions in no particular order:
> 
> - you don't use bndlib, can we remove that? bnd is fully standalone for
> your purposes and is all you need

If we don't need it then +1 to dropping it.

> - bootstrap.jar in master seems to be missing the class
> `org.apache.tomcat.util.ExceptionUtils` at least from a dependency
> analysis point of view, should it be included in the files list?

No sure. That needs some more careful thought / analysis.

> - tomcat-embed-core.jar appears to be missing tribes dependency (at
> least it refers to it but does not include it) Shouldn't it really
> contain all of catalina (minus catalina-ant?)

That is deliberate.

> - files.tomcat-embed-core seems to reference an old package
> org/apache/jk/**, can we remove that?

Yes, that can be removed.

> - I've split out the i18n LocalStrings_* (the locale specific ones) into
> a separate output dir so that you wouldn't have to then filter them out,
> it makes building their zips also cleaner, does that sound logical?

I'd want to see the patch for that first.

> - I moved a slew of concerns that bnd could handle out of ant and
> directly into the bnd configurations, manifest filtering, etc., bnd has
> filtering capability built in and knows about all ant properties so this
> eliminates ant work, comments on that?

Again, I'd want to see the patch for that.

> - generated all the Name sections for spec API packages instead of
> maintaining those by hand (now they are complete)

Sounds good.

> - add all the correct OSGi metadata of course to the spec jars and to
> everyone else.

Sounds good. Note the naming convention may have changed in the move
from Java EE to Jakarta EE.

> - I've added bnd's JPMS build support to all the modules, still builds
> on Java 8, and you get JPMS module-info almost for free (very small
> amount of configurable bits)

Generally sounds good. My only concern is consistency for the spec JARs.

> Could someone provide the build instruction I'd need to do the most
> throughout test build? I see that there's a .travis script which does
> `ant -q test 2>&1 > ant-test.log` is that really all?

Yes, assuming you have properties set to point to a valid Tomcat Native
build.

> Shall I create an SVN branch to explore this? Or would a git PR be enough?

I'd recommend tacking this one issue at a time. PRs would be fine for
that. Some should be pretty quick.

Mark

-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org

Improving OSGi/bnd build and much more...

[Raymond Auge]

Hey All (particularly Markt whom I spoke with about this months and months
ago),

Noting that bnd was already added to the build I took a crack at making
better use of it.

Here are a few items and some questions in no particular order:

- you don't use bndlib, can we remove that? bnd is fully standalone for
your purposes and is all you need
- bootstrap.jar in master seems to be missing the class
`org.apache.tomcat.util.ExceptionUtils` at least from a dependency analysis
point of view, should it be included in the files list?
- tomcat-embed-core.jar appears to be missing tribes dependency (at least
it refers to it but does not include it) Shouldn't it really contain all of
catalina (minus catalina-ant?)
- files.tomcat-embed-core seems to reference an old package
org/apache/jk/**, can we remove that?
- I've split out the i18n LocalStrings_* (the locale specific ones) into a
separate output dir so that you wouldn't have to then filter them out, it
makes building their zips also cleaner, does that sound logical?
- I moved a slew of concerns that bnd could handle out of ant and directly
into the bnd configurations, manifest filtering, etc., bnd has filtering
capability built in and knows about all ant properties so this eliminates
ant work, comments on that?
- generated all the Name sections for spec API packages instead of
maintaining those by hand (now they are complete)
- add all the correct OSGi metadata of course to the spec jars and to
everyone else.
- I've added bnd's JPMS build support to all the modules, still builds on
Java 8, and you get JPMS module-info almost for free (very small amount of
configurable bits)

Could someone provide the build instruction I'd need to do the most
throughout test build? I see that there's a .travis script which does `ant
-q test 2>&1 > ant-test.log` is that really all?

Shall I create an SVN branch to explore this? Or would a git PR be enough?
-- 
*Raymond Augé* 
 (@rotty3000)
Senior Software Architect *Liferay, Inc.* 
 (@Liferay)