Support BSD and Caciocavallo

[gnu_andrew]

This patch adds support for downloading the BSD and Caciocavallo trees
to IcedTea.
You can choose a project using --with-project and a value of jdk7 (the
default), closures,
cvmi, caciocavallo and bsd to select a tree.  It currently only makes
a difference if you also add --enable-hg; the two are separated so
additional project-specific rules can be added.  I tried building the
caciocavallo and closures trees but the IcedTea patches need some work
to apply cleanly (presumably because these trees are based on older
build drops).

2008-08-30  Andrew John Hughes  <gnu_andrew@...>

        * Makefile.am: Add BSD and Caciocavallo URLs.
        * acinclude.m4:
        (WITH_PROJECT): Added.
        (ENABLE_OTHER_PROJECTS): Converted to WITH_PROJECT.
        * configure.ac: Support building with BSD
        and Caciocavallo projects via --with-project.

Note that jdk7 is now finally at b33, just as b34 comes out...

Bedtime,
--
Andrew :-)

Support Free Java!
Contribute to GNU Classpath and the OpenJDK
http://www.gnu.org/software/classpath
http://openjdk.java.net

PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8

diff -r 40ffba722e7d ChangeLog
--- a/ChangeLog Sat Aug 30 00:14:16 2008 +0100
+++ b/ChangeLog Sat Aug 30 02:31:21 2008 +0100
@@ -1,4 +1,4 @@
-2008-08-25  Andrew John Hughes  <gnu_andrew@...>
[phone]  Andrew John Hughes  <gnu_andrew@...>
 
  * patches/icedtea-ecj.patch: Fix OGLContext.h to
  use correct headers when generated by gjavah.
diff -r 40ffba722e7d Makefile.am
--- a/Makefile.am Sat Aug 30 00:14:16 2008 +0100
+++ b/Makefile.am Sat Aug 30 02:31:21 2008 +0100
@@ -256,9 +256,10 @@
 OPENJDK_HG_URL = http://hg.openjdk.java.net/jdk7/jdk7
 CVMI_HG_URL = http://hg.openjdk.java.net/cvmi/cvmi/
 CLOSURES_HG_URL = http://hg.openjdk.java.net/closures/closures/
+CACIOCAVALLO_HG_URL = http://hg.openjdk.java.net/caciocavallo/jdk7/
+BSD_HG_URL = http://hg.openjdk.java.net/bsd-port/bsd-port
 
 OPENJDK_URL = http://www.java.net/download/openjdk/jdk7/promoted/$(OPENJDK_VERSION)/
-
 if USE_ALT_OPENJDK_SRC_ZIP
   OPENJDK_SRC_ZIP = $(ALT_OPENJDK_SRC_ZIP)
 else
@@ -296,7 +297,15 @@
 if USE_CVMI
  hg fclone $(CVMI_HG_URL) openjdk;
 else
+if USE_CACIOCAVALLO
+ hg fclone $(CACIOCAVALLO_HG_URL) openjdk;
+else
+if USE_BSD
+ hg fclone $(BSD_HG_URL) openjdk;
+else
  hg fclone -r jdk7-$(OPENJDK_VERSION) $(OPENJDK_HG_URL) openjdk;
+endif
+endif
 endif
 endif
 else
diff -r 40ffba722e7d acinclude.m4
--- a/acinclude.m4 Sat Aug 30 00:14:16 2008 +0100
+++ b/acinclude.m4 Sat Aug 30 02:31:21 2008 +0100
@@ -802,42 +802,30 @@
   AM_CONDITIONAL([USE_HG], test x"${enable_hg}" = "xyes")
 ])
 
-AC_DEFUN([ENABLE_EXTRA_PROJECTS],
+AC_DEFUN([WITH_PROJECT],
 [
-  AC_MSG_CHECKING(whether to use code from the Closures project)
-  AC_ARG_ENABLE([closures],
-                [AS_HELP_STRING(--enable-closures,use Closures source code [[default=no]])],
+  AC_MSG_CHECKING(which OpenJDK project is being used)
+  AC_ARG_WITH([project],
+              [AS_HELP_STRING(--with-project,choose the OpenJDK project to use: jdk7 closures cvmi cacioavallo bsd [[default=jdk7]])],
   [
-    case "${enableval}" in
+    case "${withval}" in
+      yes)
+ project=jdk7
+        ;;
       no)
- enable_closures=no
-        ;;
+ project=jdk7
+ ;;
       *)
-        enable_closures=yes
+        project=${withval}
         ;;
     esac
   ],
   [
-    enable_closures=no
+    project=jdk7
   ])
-  AC_MSG_RESULT([${enable_closures}])
-  AM_CONDITIONAL([USE_CLOSURES], test x"${enable_closures}" = "xyes")
-  AC_MSG_CHECKING(whether to use code from the CVMI project)
-  AC_ARG_ENABLE([cvmi],
-                [AS_HELP_STRING(--enable-cvmi,use CVMI source code [[default=no]])],
-  [
-    case "${enableval}" in
-      no)
- enable_cvmi=no
-        ;;
-      *)
-        enable_cvmi=yes
-        ;;
-    esac
-  ],
-  [
-    enable_cvmi=no
-  ])
-  AC_MSG_RESULT([${enable_cvmi}])
-  AM_CONDITIONAL([USE_CVMI], test x"${enable_cvmi}" = "xyes")
+  AC_MSG_RESULT([${project}])
+  AM_CONDITIONAL([USE_CLOSURES], test x"${project}" = "xclosures")
+  AM_CONDITIONAL([USE_CVMI], test x"${project}" = "xcvmi")
+  AM_CONDITIONAL([USE_CACIOCAVALLO], test x"${project}" = "xcaciocavallo")
+  AM_CONDITIONAL([USE_BSD], test x"${project}" = "xbsd")
 ])
diff -r 40ffba722e7d configure.ac
--- a/configure.ac Sat Aug 30 00:14:16 2008 +0100
+++ b/configure.ac Sat Aug 30 02:31:21 2008 +0100
@@ -215,7 +215,7 @@
 ENABLE_ZERO_BUILD
 SET_CORE_OR_SHARK_BUILD
 ENABLE_HG
-ENABLE_EXTRA_PROJECTS
+WITH_PROJECT
 
 if test "x${enable_hg}" = "xyes"; then
   FIND_TOOL([MERCURIAL],[hg])

[Mark Wielaard]

Hi Andrew,

On Sat, 2008-08-30 at 02:39 +0100, Andrew John Hughes wrote:
> This patch adds support for downloading the BSD and Caciocavallo trees
> to IcedTea.
> You can choose a project using --with-project and a value of jdk7 (the
> default), closures, cvmi, caciocavallo and bsd to select a tree.

Heay, that is super cool!

> It currently only makes
> a difference if you also add --enable-hg; the two are separated so
> additional project-specific rules can be added.

I immediately fell into the "--enable-hg trap" by forgetting it. I think
it makes sense to either make that the default if --with-project=x is
given and project x needs a hg repo pull, or at least fail the
configure/build step.

>   I tried building the
> caciocavallo and closures trees but the IcedTea patches need some work
> to apply cleanly (presumably because these trees are based on older
> build drops).

Yeah, this might be a bit of a pain to keep current. Either the projects
supported need to sync often enough or we need to differentiate our
patches. Neither is really that attractive. The best thing would
probably be for us to help these projects get pushed into the main jdk7
tree sooner.

> 2008-08-30  Andrew John Hughes  <gnu_andrew@...>
>
> * Makefile.am: Add BSD and Caciocavallo URLs.
> * acinclude.m4:
> (WITH_PROJECT): Added.
> (ENABLE_OTHER_PROJECTS): Converted to WITH_PROJECT.
> * configure.ac: Support building with BSD
> and Caciocavallo projects via --with-project.
>
> Note that jdk7 is now finally at b33, just as b34 comes out...

You managed to push this without generating a commit message. I cannot
figure out why that is. Do you have any idea how you got this commit in
"under the hgext.notify radar"?

Cheers,

Mark

[gnu_andrew]

On 31/08/2008, Mark Wielaard <mark@...> wrote:

> Hi Andrew,
>
>
>  On Sat, 2008-08-30 at 02:39 +0100, Andrew John Hughes wrote:
>  > This patch adds support for downloading the BSD and Caciocavallo trees
>  > to IcedTea.
>  > You can choose a project using --with-project and a value of jdk7 (the
>  > default), closures, cvmi, caciocavallo and bsd to select a tree.
>
>
> Heay, that is super cool!
>
>

:)

>  > It currently only makes
>  > a difference if you also add --enable-hg; the two are separated so
>  > additional project-specific rules can be added.
>
>
> I immediately fell into the "--enable-hg trap" by forgetting it. I think
>  it makes sense to either make that the default if --with-project=x is
>  given and project x needs a hg repo pull, or at least fail the
>  configure/build step.
>

Attached patch fixes this:

2008-08-31  Andrew John Hughes  <gnu_andrew@...>

        * Makefile.am:
        Make enabling hg-only projects sufficient for
        downloading from hg.

This does mean I had to separate out the hg forest check so it wasn't
duplicated.

>
>  >   I tried building the
>  > caciocavallo and closures trees but the IcedTea patches need some work
>  > to apply cleanly (presumably because these trees are based on older
>  > build drops).
>
>
> Yeah, this might be a bit of a pain to keep current. Either the projects
>  supported need to sync often enough or we need to differentiate our
>  patches. Neither is really that attractive. The best thing would
>  probably be for us to help these projects get pushed into the main jdk7
>  tree sooner.
>

Some patches are more prone to this than others, or, in other words, there are
some patches that break with just about every build drop as well.
I'll try and clean
some of them up so they are less prone to this.  Notably, I think most
of patches/icedtea-copy-plugs.patch should be part of
icedtea-ecj.patch and the rest should be dropped.  It would also be
nice to split up icedtea-ecj as it covers a lot of issues at present,
for which the only relationship is that they fix breakages that only
occur with an ecj + GNU Classpath build.

>
>  > 2008-08-30  Andrew John Hughes  <gnu_andrew@...>
>  >
>  >       * Makefile.am: Add BSD and Caciocavallo URLs.
>  >       * acinclude.m4:
>  >       (WITH_PROJECT): Added.
>  >       (ENABLE_OTHER_PROJECTS): Converted to WITH_PROJECT.
>  >       * configure.ac: Support building with BSD
>  >       and Caciocavallo projects via --with-project.
>  >
>  > Note that jdk7 is now finally at b33, just as b34 comes out...
>
>
> You managed to push this without generating a commit message. I cannot
>  figure out why that is. Do you have any idea how you got this commit in
>  "under the hgext.notify radar"?
>

Something is broken:

remote: added 1 changesets with 2 changes to 2 files
remote: error: incoming.notify hook raised an exception: cannot
concatenate 'str' and 'NoneType' objects

But at least it means merges are quiet ;)

>  Cheers,
>
>
>  Mark
>
>


--
Andrew :-)

Support Free Java!
Contribute to GNU Classpath and the OpenJDK
http://www.gnu.org/software/classpath
http://openjdk.java.net

PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8

diff -r 8d00b91f5225 Makefile.am
--- a/Makefile.am Sat Aug 30 02:33:44 2008 +0100
+++ b/Makefile.am Sun Aug 31 20:55:39 2008 +0100
@@ -280,17 +280,10 @@
 endif
 
 
-stamps/download.stamp:
+stamps/download.stamp: stamps/hgforest.stamp
 if !OPENJDK_SRC_DIR_FOUND
 if USE_ALT_OPENJDK_SRC_ZIP
 else
-if USE_HG
- if ! hg fclone -h; \
-          then \
-          echo "No forest extension found. Please refer to" \
-            "http://selenic.com/mercurial/wiki/index.cgi/ForestExtension"; \
-          exit 1; \
-          fi;
 if USE_CLOSURES
  hg fclone $(CLOSURES_HG_URL) openjdk;
 else
@@ -303,11 +296,8 @@
 if USE_BSD
  hg fclone $(BSD_HG_URL) openjdk;
 else
+if USE_HG
  hg fclone -r jdk7-$(OPENJDK_VERSION) $(OPENJDK_HG_URL) openjdk;
-endif
-endif
-endif
-endif
 else
  if ! echo "$(OPENJDK_MD5SUM)  $(OPENJDK_SRC_ZIP)" \
   | $(MD5SUM) --check ; \
@@ -318,6 +308,10 @@
   fi ; \
   $(WGET) $(OPENJDK_URL)$(OPENJDK_SRC_ZIP) -O $(OPENJDK_SRC_ZIP); \
  fi
+endif
+endif
+endif
+endif
 endif
 endif
 endif
@@ -339,6 +333,36 @@
 endif
  mkdir -p stamps
  touch stamps/download.stamp
+
+stamps/hgforest.stamp:
+if OPENJDK_SRC_DIR_FOUND
+else
+if USE_ALT_OPENJDK_SRC_ZIP
+else
+if !USE_HG
+else
+if !USE_CLOSURES
+else
+if !USE_CVMI
+else
+if !USE_CACIOCAVALLO
+else
+if !USE_BSD
+else
+ if ! hg fclone -h; \
+          then \
+          echo "No forest extension found. Please refer to" \
+            "http://selenic.com/mercurial/wiki/index.cgi/ForestExtension"; \
+          exit 1; \
+          fi;
+endif
+endif
+endif
+endif
+endif
+endif
+endif
+ touch stamps/hgforest.stamp
 
 clean-extract:
  rm -f stamps/extract.stamp

[C. K. Jester-Young]

Hi Andrew,

On Sun, Aug 31, 2008 at 09:01:56PM +0100, Andrew John Hughes wrote:
> 2008-08-31  Andrew John Hughes  <gnu_andrew@...>
>
> * Makefile.am:
> Make enabling hg-only projects sufficient for
> downloading from hg.
>
> This does mean I had to separate out the hg forest check so it wasn't
> duplicated.

I like the idea of the fix, though I think the way it's implemented
doesn't do what's intended, and is additionally a maintenance hassle
whenever new Mercurial repositories are added.

Looking at the forest test, it only applies if !OPENJDK_SRC_DIR_FOUND
&& !USE_ALT_OPENJDK_SRC_ZIP && USE_HG && USE_CLOSURES && USE_CVMI &&
USE_CACIOCAVALLO && USE_BSD. In other words, this test will never be
invoked.

I propose an alternative patch, that simply turns --enable-hg on if
--with-project != jdk7. The main caveat is now that the --with-project
stuff must appear before the --enable-hg. But other than that, I think
that this is more robust and easier to maintain, and I hope it works
for you too. :-)

Many thanks,
        ---Chris K.

diff --git a/acinclude.m4 b/acinclude.m4
--- a/acinclude.m4
+++ b/acinclude.m4
@@ -796,7 +796,14 @@ AC_DEFUN([ENABLE_HG],
     esac
   ],
   [
-    enable_hg=no
+    case "${project}" in
+      jdk7)
+        enable_hg=no
+        ;;
+      *)
+        enable_hg=yes
+        ;;
+    esac
   ])
   AC_MSG_RESULT([${enable_hg}])
   AM_CONDITIONAL([USE_HG], test x"${enable_hg}" = "xyes")
diff --git a/configure.ac b/configure.ac
--- a/configure.ac
+++ b/configure.ac
@@ -214,8 +214,8 @@ ENABLE_OPTIMIZATIONS
 ENABLE_OPTIMIZATIONS
 ENABLE_ZERO_BUILD
 SET_CORE_OR_SHARK_BUILD
+WITH_PROJECT
 ENABLE_HG
-WITH_PROJECT
 
 if test "x${enable_hg}" = "xyes"; then
   FIND_TOOL([MERCURIAL],[hg])

[Mark Wielaard]

On Sun, 2008-08-31 at 21:01 +0100, Andrew John Hughes wrote:
> On 31/08/2008, Mark Wielaard <mark@...> wrote:
> > You managed to push this without generating a commit message. I cannot
> >  figure out why that is. Do you have any idea how you got this commit in
> >  "under the hgext.notify radar"?

> Something is broken:
>
> remote: added 1 changesets with 2 changes to 2 files
> remote: error: incoming.notify hook raised an exception: cannot
> concatenate 'str' and 'NoneType' objects
>
> But at least it means merges are quiet ;)

Yeah... But this is slightly worrying.
I can replicate it when pushing remotely, but only with the icedtea6
repo, not with for example the testrepo (when enabling mail
notifications on that one). And when pushing locally I don't get any
errors which makes debugging this issue somewhat difficult.

The only hint from a quick search was not having diffstat installed
(server has hg 0.9.3 installed). But that isn't it because diffstat is,
and always was, installed.

hmmm,

Mark

[gnu_andrew]

On 01/09/2008, C. K. Jester-Young <cky944@...> wrote:

> Hi Andrew,
>
>
>  On Sun, Aug 31, 2008 at 09:01:56PM +0100, Andrew John Hughes wrote:
>  > 2008-08-31  Andrew John Hughes  <gnu_andrew@...>
>  >
>  >       * Makefile.am:
>  >       Make enabling hg-only projects sufficient for
>  >       downloading from hg.
>  >
>  > This does mean I had to separate out the hg forest check so it wasn't
>  > duplicated.
>
>
> I like the idea of the fix, though I think the way it's implemented
>  doesn't do what's intended, and is additionally a maintenance hassle
>  whenever new Mercurial repositories are added.
>
>  Looking at the forest test, it only applies if !OPENJDK_SRC_DIR_FOUND
>  && !USE_ALT_OPENJDK_SRC_ZIP && USE_HG && USE_CLOSURES && USE_CVMI &&
>  USE_CACIOCAVALLO && USE_BSD. In other words, this test will never be
>  invoked.
>
>  I propose an alternative patch, that simply turns --enable-hg on if
>  --with-project != jdk7. The main caveat is now that the --with-project
>  stuff must appear before the --enable-hg. But other than that, I think
>  that this is more robust and easier to maintain, and I hope it works
>  for you too. :-)
>
>  Many thanks,
>         ---Chris K.
>
>

This does simplify it quite a bit.  Applying with the following
addition from myself:

2008-09-01  Andrew John Hughes  <gnu_andrew@...>

        * Makefile.am: Simply hgforest
        rule and add corresponding clean rule.

2008-09-01  C. K. Jester-Young  <cky944@...>
       
        * acinclude.m4:
        Enable hg by default unless building jdk7.
        * configure.ac:
        Do project check before hg check.

--
Andrew :-)

Support Free Java!
Contribute to GNU Classpath and the OpenJDK
http://www.gnu.org/software/classpath
http://openjdk.java.net

PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8

diff -r a569986e9839 Makefile.am
--- a/Makefile.am Mon Sep 01 14:41:43 2008 +0100
+++ b/Makefile.am Mon Sep 01 15:01:40 2008 +0100
@@ -415,14 +415,6 @@
 else
 if !USE_HG
 else
-if !USE_CLOSURES
-else
-if !USE_CVMI
-else
-if !USE_CACIOCAVALLO
-else
-if !USE_BSD
-else
  if ! hg fclone -h; \
           then \
           echo "No forest extension found. Please refer to" \
@@ -432,11 +424,10 @@
 endif
 endif
 endif
-endif
-endif
-endif
-endif
  touch stamps/hgforest.stamp
+
+clean-hgforest:
+ rm -f stamps/hgforest.stamp
 
 clean-extract:
  rm -f stamps/extract.stamp
diff -r a569986e9839 acinclude.m4
--- a/acinclude.m4 Mon Sep 01 14:41:43 2008 +0100
+++ b/acinclude.m4 Mon Sep 01 15:01:40 2008 +0100
@@ -878,7 +878,14 @@
     esac
   ],
   [
-    enable_hg=no
+    case "${project}" in
+      jdk7)
+        enable_hg=no
+        ;;
+      *)
+        enable_hg=yes
+        ;;
+    esac
   ])
   AC_MSG_RESULT([${enable_hg}])
   AM_CONDITIONAL([USE_HG], test x"${enable_hg}" = "xyes")
diff -r a569986e9839 configure.ac
--- a/configure.ac Mon Sep 01 14:41:43 2008 +0100
+++ b/configure.ac Mon Sep 01 15:01:40 2008 +0100
@@ -225,8 +225,8 @@
 ENABLE_OPTIMIZATIONS
 ENABLE_ZERO_BUILD
 SET_CORE_OR_SHARK_BUILD
+WITH_PROJECT
 ENABLE_HG
-WITH_PROJECT
 
 if test "x${enable_hg}" = "xyes"; then
   FIND_TOOL([MERCURIAL],[hg])

[Mario Torre-2]

Il giorno dom, 31/08/2008 alle 21.01 +0100, Andrew John Hughes ha
scritto:

> On 31/08/2008, Mark Wielaard <mark@...> wrote:
> > Hi Andrew,
> >
> >
> >  On Sat, 2008-08-30 at 02:39 +0100, Andrew John Hughes wrote:
> >  > This patch adds support for downloading the BSD and Caciocavallo trees
> >  > to IcedTea.
> >  > You can choose a project using --with-project and a value of jdk7 (the
> >  > default), closures, cvmi, caciocavallo and bsd to select a tree.
> >
> >
> > Heay, that is super cool!

Hi Andrew!

Hey I missed this, thanks!

> >
> > Yeah, this might be a bit of a pain to keep current. Either the projects
> >  supported need to sync often enough or we need to differentiate our
> >  patches. Neither is really that attractive. The best thing would
> >  probably be for us to help these projects get pushed into the main jdk7
> >  tree sooner.
> >
>
> Some patches are more prone to this than others, or, in other words, there are
> some patches that break with just about every build drop as well.
> I'll try and clean

We have few patches in our cacio repo that are quite sensible to this,
too, most notably the Font related patches.

We need these in a couple of places, and we should discuss this with the
Sun guys and find a good solution for the Font stuff because our patches
there are hacky, but the current code is not much better, and this is
one of the areas where the peers code is quite sensible.

We are going on with the project anyway, and I think we will port the
new stuff to a newer jdk drop in a couple of weeks or so
(optimistically, there is always no real time for doing anything), so I
think that the icedtea patches will apply cleanly again.

I also think that we had to "steal" a couple of patches from icedtea,
probably the ones related to gcc 4.3, so one reason why they don't apply
cleanly maybe because they are already there?

I'll try to give it a look this weekend.

Thanks,
Mario

[Mark Wielaard]

Hi Andrew,

On Sun, 2008-08-31 at 21:01 +0100, Andrew John Hughes wrote:

> On 31/08/2008, Mark Wielaard <mark@...> wrote:
> > You managed to push this without generating a commit message. I cannot
> >  figure out why that is. Do you have any idea how you got this commit in
> >  "under the hgext.notify radar"?
> >
> Something is broken:
>
> remote: added 1 changesets with 2 changes to 2 files
> remote: error: incoming.notify hook raised an exception: cannot
> concatenate 'str' and 'NoneType' objects

I tried to track this down, but it seems things are just working fine
now. I am slightly concerned that this is intermittent and we might be
missing commit emails. Could you let me know if you get this error
again?

Thanks,

Mark

[Mark Wielaard]

Hi all,

On Thu, 2008-09-25 at 14:45 +0200, Mark Wielaard wrote:

> On Sun, 2008-08-31 at 21:01 +0100, Andrew John Hughes wrote:
> > On 31/08/2008, Mark Wielaard <mark@...> wrote:
> > > You managed to push this without generating a commit message. I cannot
> > >  figure out why that is. Do you have any idea how you got this commit in
> > >  "under the hgext.notify radar"?
> > >
> > Something is broken:
> >
> > remote: added 1 changesets with 2 changes to 2 files
> > remote: error: incoming.notify hook raised an exception: cannot
> > concatenate 'str' and 'NoneType' objects

We finally found it!
The issue is when your hgrc doesn't contain an email address.
So please make sure that your ~/.hgrc contains at least the following:

### --- User interface
[ui]
### username data to appear in comits
### it usually takes the form: Joe User <joe.user@...>
username = My FullName <me@...>

If for some reason you need your global ~/.hgrc to contain something
else than your full name then add this to your project specific hgrc:
  icedtea/.hg/hgrc
  icedtea6/.hg/hgrc
  icepick/.hg/hgrc
  pulseaudio/.hg/hgrc
etc.

That way your commits will always have a full name plus the email
address added. Which will then be used to send to notification emails to
the mailinglist.

Thanks,

Mark

[Kelly O'Hair]

Humm...

Anything in the [ui] username will be put into any changesets created
by that user (e.g. hg commit or hg tag), verbatim, as a permanent record.
In the OpenJDK project we insisted on people setting username to the
OpenJDK name that you registered as, and ONLY that name.
We were concerned about email addresses getting buried in the changesets
and providing too much of a source of valid email addresses for spambots.

Does the [email] settings not work? e.g.

[email]
from = your@...

Seems like it should.

-kto

Mark Wielaard wrote:

> Hi all,
>
> On Thu, 2008-09-25 at 14:45 +0200, Mark Wielaard wrote:
>> On Sun, 2008-08-31 at 21:01 +0100, Andrew John Hughes wrote:
>>> On 31/08/2008, Mark Wielaard <mark@...> wrote:
>>>> You managed to push this without generating a commit message. I cannot
>>>>  figure out why that is. Do you have any idea how you got this commit in
>>>>  "under the hgext.notify radar"?
>>>>
>>> Something is broken:
>>>
>>> remote: added 1 changesets with 2 changes to 2 files
>>> remote: error: incoming.notify hook raised an exception: cannot
>>> concatenate 'str' and 'NoneType' objects
>
> We finally found it!
> The issue is when your hgrc doesn't contain an email address.
> So please make sure that your ~/.hgrc contains at least the following:
>
> ### --- User interface
> [ui]
> ### username data to appear in comits
> ### it usually takes the form: Joe User <joe.user@...>
> username = My FullName <me@...>
>
> If for some reason you need your global ~/.hgrc to contain something
> else than your full name then add this to your project specific hgrc:
>   icedtea/.hg/hgrc
>   icedtea6/.hg/hgrc
>   icepick/.hg/hgrc
>   pulseaudio/.hg/hgrc
> etc.
>
> That way your commits will always have a full name plus the email
> address added. Which will then be used to send to notification emails to
> the mailinglist.
>
> Thanks,
>
> Mark
>

[gnu_andrew]

2008/9/27 Mark Wielaard <mark@...>:

> Hi all,
>
> On Thu, 2008-09-25 at 14:45 +0200, Mark Wielaard wrote:
>> On Sun, 2008-08-31 at 21:01 +0100, Andrew John Hughes wrote:
>> > On 31/08/2008, Mark Wielaard <mark@...> wrote:
>> > > You managed to push this without generating a commit message. I cannot
>> > >  figure out why that is. Do you have any idea how you got this commit in
>> > >  "under the hgext.notify radar"?
>> > >
>> > Something is broken:
>> >
>> > remote: added 1 changesets with 2 changes to 2 files
>> > remote: error: incoming.notify hook raised an exception: cannot
>> > concatenate 'str' and 'NoneType' objects
>
> We finally found it!
> The issue is when your hgrc doesn't contain an email address.
> So please make sure that your ~/.hgrc contains at least the following:
>
> ### --- User interface
> [ui]
> ### username data to appear in comits
> ### it usually takes the form: Joe User <joe.user@...>
> username = My FullName <me@...>
>
> If for some reason you need your global ~/.hgrc to contain something
> else than your full name then add this to your project specific hgrc:
>  icedtea/.hg/hgrc
>  icedtea6/.hg/hgrc
>  icepick/.hg/hgrc
>  pulseaudio/.hg/hgrc
> etc.
>
> That way your commits will always have a full name plus the email
> address added. Which will then be used to send to notification emails to
> the mailinglist.
>
> Thanks,
>
> Mark
>
>

Yes as you'll have noticed my commits now come through since setting this
in icedtea6/.hg/hgrc.
--
Andrew :-)

Support Free Java!
Contribute to GNU Classpath and the OpenJDK
http://www.gnu.org/software/classpath
http://openjdk.java.net

PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8

[Mark Wielaard]

Hi Kelly,

On Sat, 2008-09-27 at 13:27 -0700, Kelly O'Hair wrote:
> Anything in the [ui] username will be put into any changesets created
> by that user (e.g. hg commit or hg tag), verbatim, as a permanent record.
> In the OpenJDK project we insisted on people setting username to the
> OpenJDK name that you registered as, and ONLY that name.

I assumed this was an old leftover from the time that development was
centralized. Now that development is done in the open, with multiple
parties and using a distributed source code management system it seems
somewhat odd to introduce a naming scheme for developers that is
centralized again. Also since there is no ChangeLog file in the various
openjdk repositories it is harder to see who really made a change if
someone isn't using their full name and email address. Personally I have
found it pretty hard to only have these handles without any other
identifying information in the commit messages. People do have names :)
And having contact information for changesets is a good thing imho.

> Does the [email] settings not work? e.g.
>
> [email]
> from = your@...
>
> Seems like it should.

Maybe we can get the notify extension to use that locally. But that
still doesn't give you the full name and email address of the person
actually having done the commit into the repository that the extension
is monitoring. If the information isn't put into the commit message it
is essentially lost and cannot easily be recovered when moving patch
sets between repos.

Cheers,

Mark