Community
Participate
Working Groups
Using eclipse-SDK-I20191112-1800-win32-x86_64.zip, it seems that the recent changes to workspace selection have opened a hole. To start a new workspace, I took my usual short and changed the workspace to the required new name. Starting Eclipse via the new short cut crashes; giving the Java failed pop up. Surely it should just work, or perhaps offer a do you want to create workspace pop up? Certainly not a crash.
(In reply to Ed Willink from comment #0) > Using eclipse-SDK-I20191112-1800-win32-x86_64.zip, it seems that the recent > changes to workspace selection have opened a hole. *which* changes? > To start a new workspace, I took my usual short and changed the workspace to > the required new name. Starting Eclipse via the new short cut crashes; > giving the Java failed pop up. Please provide error log and clear steps to reproduce. > Surely it should just work, or perhaps offer a do you want to create > workspace pop up? Certainly not a crash. If it crashed, please provide crash dump.
The actual problem in the re-used short cut E:\Tools\Eclipse\4.11\eclipse.exe -clean -data E:\Development\Rusa\Workspace -vm "%JAVA_HOME%\bin\javaw.exe" -vmargs -Xmx4g -XX:PermSize=100M -XX:MaxPermSize=300M -Xrunjdwp:server=y,transport=dt_socket,address=14143,suspend=n was not the new workspace but the re-used 14143 socket. So behind the scenes was some form of port-in-use IOException. (In reply to Andrey Loskutov from comment #1) > If it crashed, please provide crash dump. Unfortunately the Eclipse-failed-to-launch dialog is pretty verbose and also pretty useless. It shows the full command line (sometimes useful) but no more diagnosis than Java exited with some numeric code. The IOException does not appear directly. There is no hyperlink to a log file that might contain it. There is no hyperlink to a crash dump. ---- Therefore this bug is the absence of any clues that the failure to launch with a re-used socket was due to a socket connection failure.
Closing as user error. VM can't start if the debug port is already in use.
Yes. But the Eclipse launcher runs before the VM start so surely the Eclipse launcher can diagnose the problem much much better?
(In reply to Ed Willink from comment #4) > Yes. But the Eclipse launcher runs before the VM start so surely the Eclipse > launcher can diagnose the problem much much better? In theory, yes. However it will require parsing of various ways to open some port (like -Xrunjdwp) and checking if that port is in use on the current system, which is probably not worth the effort. If you have time & energy to contribute this non trivial, cross-platform VM attributes parsing/network checking C++ code (launcher code is C++), feel free to reopen.
Surely the VM has an API like Process; simple programming just propagates the trivial integer exitValue(); serious applications hook up the output/errorStream? And as a minimum the dialog could identify the directory of a log file / dump.
https://wiki.eclipse.org/Platform_UI/How_to_Contribute
You may not have noticed but I in addition to three existing projects, I am just making another major contribution by rescuing for Modisco from termination. I do not have time to get up to speed on the platform whose developers can react at least ten times more proficiently than me.
Ed, I've just added a hint how to proceed here (and also did it in my free time).
