David Crossley wrote:
David Crossley wrote:
David Crossley wrote:
David Crossley wrote:
David Crossley wrote:
I notice that there is a different size with the jar
that gets packed for the release ...
[forrest-trunk-release]$ ls -l build/xml-forrest.jar build/dist/shbat/build/xml-forrest.jar
67164 Apr 10 22:34 build/dist/shbat/build/xml-forrest.jar
66577 Apr 10 22:33 build/xml-forrest.jar
This is very strange. All that happened in the "dist-shbat"
target is that the Ant "copy" task copied the build/xml-forrest.jar
to that dist location.
Ah, i made the "dist-shbat" target pause with an "input"
task, copied the jar over by hand, then let it proceed
after the copy.
Something mangles the jar later in the process.
Looking ...
Found it.
I had added "fixcrlf" tasks so that we didn't need
to build the release on two different operating systems.
It was wrecking the jar file.
Doing 'forrest run' works for me now.
However, there are some images that are mangled too.
I thought that "fixcrlf" would only touch text files,
but no.
So i think that we need to go back to how we built
the packages in the past.
Would a committer on Windows (preferably have a well-connected
PGP key) please get ready to build the release.
Get a fresh trunk checkout and have Java 1.4
I will help you with the next steps.
Actually hold that.
There is a reasonably small set of binary filename
extensions. I could try excluding them too from the
fixcrlf task.
Surely other projects don't need to jump through
such hoops. Do they worry about line-endings?
I am tempted to just build the first release
candidate without the fixcrlf and see what happens.
Does anyone have an opinion on that? Need to make
a decision quickly.
I'm not in a position to build windows version on the machine I am
currently using as I can't put JDK1.4 on it. However, I can test (and
quickly) a build you do on this machine (windows XP Pro, JDK 1.5)
I can put JDK 1.4 on my home machine but I've given that to my wife so
not sure when I can lay my hands on it.
Ross
