Francois Orsini wrote: >Hi kathey, > >Am guessing (and as you did) that the java Runtime.exec() (called by ant) >ultimately calls Win32 CreateProcess() and causes this last one to fail as >the path to the md5.ksh script is in a Unix form (/marsden/bin/md5.ksh) - >Hence Win32 CreateProcess() cannot deal with Cygwin path conventions and >would not be able to locate it... > >Now what's puzzling is why the MD5 hash differs between the 2 different >invocation modes... > > > Well I think the slashes got turned around as my error had them in the DOS format but probably some other problem like the fact that there is a pipe in the script.
java.io.IOException: CreateProcess: \marsden\bin\md5.ksh D:\svn\opensource\10.1/tools/release/db-derby-10.1.2.0-bin.zip error=193 For the hash differences I think I must have picked up the output from two different attempts. It does match up, but the prepended '\' seems perhaps troublesome too. $ c:/build/sdks/cygnustools/md5sum 'D:\svn\opensource\10.1/tools/release/db- \70b4807b544384d16db0ff0a0a299dc0 *D:\\svn\\opensource\\10.1/tools/release/db-derby-10.1.2.0-bin.zip $ md5.ksh 'D:\svn\opensource\10.1/tools/release/db-derby-10.1.2.0-bin.zip' \70b4807b544384d16db0ff0a0a299dc0 I also am not sure of the exact format of md5 -q that I am supposed to be matching. So, since I am the Dummy in this Releases for Dummies excercise, Andrew would you mind fixing up the doc to give an example of how the packaging.properties can be modified to work on Windows? Thanks Kathey
