You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Justin Erenkrantz <ju...@erenkrantz.com> on 2004/04/28 02:11:47 UTC
Busted javahl build on Mac OS X was Re: svn commit: r9415 - in
trunk: build/ac-macros subversion/bindings/java
--On Saturday, April 17, 2004 9:45 PM -0500 dlr@tigris.org wrote:
> Author: dlr
> Date: Sat Apr 17 21:44:55 2004
> New Revision: 9415
>
> Modified:
> trunk/build/ac-macros/java.m4
> trunk/subversion/bindings/java/README
> Log:
> Improve location of the jikes Java compiler by allowing for a
> --with-jikes=[PATH|yes|no] argument to configure.
Heh. This borks the 'javahl' build on Mac OS X for me. Here I was
thinking I never finished a corner case, but then I hunted through java.m4
and saw this jikes stuff going, 'Gee, I didn't write this!' - sure enough
adding
'--without-jikes' fixes it for me.
In short, jikes isn't smart enough to find the classes by itself, and the
JAVA_CLASSPATH detection code isn't clever enough on Mac OS X. I don't
know how to programmatically find the classes.jar that jikes need (it's in
'/System/Library/Frameworks/JavaVM.framework/Classes/classes.jar' here on
my box). The default javac doesn't need to be told the classes.jar path,
so that's better, I think.
So, I think we should flip the default to not using jikes - at least on Mac
OS X.
What do you think? -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Busted javahl build on Mac OS X was Re: svn commit: r9415 - in
trunk: build/ac-macros subversion/bindings/java
Posted by Patrick Mayweg <ma...@qint.de>.
Hi Justin,
Justin Erenkrantz wrote:
> --On Saturday, April 17, 2004 9:45 PM -0500 dlr@tigris.org wrote:
>
>> Author: dlr
>> Date: Sat Apr 17 21:44:55 2004
>> New Revision: 9415
>>
>> Modified:
>> trunk/build/ac-macros/java.m4
>> trunk/subversion/bindings/java/README
>> Log:
>> Improve location of the jikes Java compiler by allowing for a
>> --with-jikes=[PATH|yes|no] argument to configure.
>
>
> Heh. This borks the 'javahl' build on Mac OS X for me. Here I was
> thinking I never finished a corner case, but then I hunted through
> java.m4 and saw this jikes stuff going, 'Gee, I didn't write this!' -
> sure enough adding
> '--without-jikes' fixes it for me.
>
> In short, jikes isn't smart enough to find the classes by itself, and
> the JAVA_CLASSPATH detection code isn't clever enough on Mac OS X. I
> don't know how to programmatically find the classes.jar that jikes
> need (it's in
> '/System/Library/Frameworks/JavaVM.framework/Classes/classes.jar' here
> on my box). The default javac doesn't need to be told the classes.jar
> path, so that's better, I think.
>
> So, I think we should flip the default to not using jikes - at least
> on Mac OS X.
>
> What do you think? -- justin
>
Since I have never used jikes before. I am fine with having javac as the
default compiler.
Patrick
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Busted javahl build on Mac OS X was Re: svn commit: r9415 - in
trunk: build/ac-macros subversion/bindings/java
Posted by "Daniel L. Rall" <dl...@finemaltcoding.com>.
Daniel Rall wrote:
> Justin Erenkrantz wrote:
...
>> Well, the problem I have is that I really dislike hardcoding the
>> JRE_LIB_DIR. If we could find a way to determine this from having the
>> JDK in /Library/Java/ or something, I'd be more comfortable. But, the
>> degree to which we're hardcoding paths is getting uncomfortable. --
>> justin
>
> Justin, I've never heard of any way to programmatically determine this
> information. (Doesn't mean there isn't one, of course.) I don't see
> this as a significant issue, though, since the highest percentage of
> people compiling Subversion will be using the JDK from Sun (or Apple, if
> on OS X ;). General path structure for the JDK's of these vendors tend
> not to vary much across versions. Sun's JDK always includes a JRE at a
> specific location relative to the root of that JDK, making this
> something of a non-issue from the get-go. Apple doesn't seem to follow
> this structure, but is likely to always install the JRE in the same
> place anyhow. If you hear of any OS X-foo for requesting the location
> of the JRE, let me know and I'll add that logic to java.m4.
>
> That said, does that patch look like it will fix the immediate problem?
> Like I said, I'm not up and running on moof yet. Did you have a chance
> to try it out?
Another option is the addition of a --with-jre configure flag which overrides
the default, (for Sun, semi-)hard-coded paths, allowing building by users with
irregular installations.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Busted javahl build on Mac OS X was Re: svn commit: r9415 - in
trunk: build/ac-macros subversion/bindings/java
Posted by Daniel Rall <dl...@collab.net>.
Justin Erenkrantz wrote:
> --On Thursday, April 29, 2004 9:36 AM -0700 "Daniel L. Rall"
> <dl...@finemaltcoding.com> wrote:
>
>> If ya got jikes, prefer it. Last I checked it was an order of magnitude
>> faster than Sun's javac, and very reliable. I have heard that Apple's
>> javac is somewhat faster than Sun's. In any case, how about something
>> along the lines of the attached patch for correctly adjusting the
>> JAVA_CLASSPATH? Untested, as I haven't gotten a build going on moof yet
>> (hit too many snags last night due to unfamiliarity with OS X and moof).
>
> Well, the problem I have is that I really dislike hardcoding the
> JRE_LIB_DIR. If we could find a way to determine this from having the
> JDK in /Library/Java/ or something, I'd be more comfortable. But, the
> degree to which we're hardcoding paths is getting uncomfortable. -- justin
Justin, I've never heard of any way to programmatically determine this
information. (Doesn't mean there isn't one, of course.) I don't see this as a
significant issue, though, since the highest percentage of people compiling
Subversion will be using the JDK from Sun (or Apple, if on OS X ;). General
path structure for the JDK's of these vendors tend not to vary much across
versions. Sun's JDK always includes a JRE at a specific location relative to
the root of that JDK, making this something of a non-issue from the get-go.
Apple doesn't seem to follow this structure, but is likely to always install the
JRE in the same place anyhow. If you hear of any OS X-foo for requesting the
location of the JRE, let me know and I'll add that logic to java.m4.
That said, does that patch look like it will fix the immediate problem? Like I
said, I'm not up and running on moof yet. Did you have a chance to try it out?
- Dan
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Busted javahl build on Mac OS X was Re: svn commit: r9415 - in
trunk: build/ac-macros subversion/bindings/java
Posted by Justin Erenkrantz <ju...@erenkrantz.com>.
--On Thursday, April 29, 2004 9:36 AM -0700 "Daniel L. Rall"
<dl...@finemaltcoding.com> wrote:
> If ya got jikes, prefer it. Last I checked it was an order of magnitude
> faster than Sun's javac, and very reliable. I have heard that Apple's
> javac is somewhat faster than Sun's. In any case, how about something
> along the lines of the attached patch for correctly adjusting the
> JAVA_CLASSPATH? Untested, as I haven't gotten a build going on moof yet
> (hit too many snags last night due to unfamiliarity with OS X and moof).
Well, the problem I have is that I really dislike hardcoding the
JRE_LIB_DIR. If we could find a way to determine this from having the JDK
in /Library/Java/ or something, I'd be more comfortable. But, the degree
to which we're hardcoding paths is getting uncomfortable. -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Busted javahl build on Mac OS X was Re: svn commit: r9415 - in
trunk: build/ac-macros subversion/bindings/java
Posted by "Daniel L. Rall" <dl...@finemaltcoding.com>.
Justin Erenkrantz wrote:
> --On Saturday, April 17, 2004 9:45 PM -0500 dlr@tigris.org wrote:
>
>> Author: dlr
>> Date: Sat Apr 17 21:44:55 2004
>> New Revision: 9415
>>
>> Modified:
>> trunk/build/ac-macros/java.m4
>> trunk/subversion/bindings/java/README
>> Log:
>> Improve location of the jikes Java compiler by allowing for a
>> --with-jikes=[PATH|yes|no] argument to configure.
>
>
> Heh. This borks the 'javahl' build on Mac OS X for me. Here I was
> thinking I never finished a corner case, but then I hunted through
> java.m4 and saw this jikes stuff going, 'Gee, I didn't write this!' -
> sure enough adding
> '--without-jikes' fixes it for me.
>
> In short, jikes isn't smart enough to find the classes by itself, and
> the JAVA_CLASSPATH detection code isn't clever enough on Mac OS X. I
> don't know how to programmatically find the classes.jar that jikes need
> (it's in
> '/System/Library/Frameworks/JavaVM.framework/Classes/classes.jar' here
> on my box). The default javac doesn't need to be told the classes.jar
> path, so that's better, I think.
>
> So, I think we should flip the default to not using jikes - at least on
> Mac OS X.
>
> What do you think? -- justin
If ya got jikes, prefer it. Last I checked it was an order of magnitude
faster than Sun's javac, and very reliable. I have heard that Apple's javac
is somewhat faster than Sun's. In any case, how about something along the
lines of the attached patch for correctly adjusting the JAVA_CLASSPATH?
Untested, as I haven't gotten a build going on moof yet (hit too many snags
last night due to unfamiliarity with OS X and moof).
- Dan