Unfortunately I could not find a workaround to this. Played with the paths trying to fool DepsGenerator but it didn't work.
The other approach is to copy the DepsGenerator and PathUtil classes to FalconJX code base and fix it there. Technically this approach works, I tried it. The fix to PathUtil is trivial to do, only a few lines must be modified to make it handle Windows paths correctly. I did not need to touch DepsGenerator except to change the import to use 'our' fixed PathUtil. After the fix is applied compiling FlexJSTest_again works and produces correct deps.js on Windows and the example then runs fine in the browser. However I do not know what are full implications and if you practice such approach anyway. Looks like OK to me from legal point of view since the Closure Compiler is Apache licensed. Please let me know if what I am doing makes any sense or I should scrap what I did so far and wait for someone smarter than me to work on this. Is there any ToDo list for FalconJX so that I can go ahead and try to work on some outstanding issues? I now have Falcon projects in my Eclipse, git is configured and am now ready for a bit more serious work. Tigran. -----Original Message----- From: Erik de Bruin [mailto:e...@ixsoftware.nl] Sent: Monday, April 08, 2013 4:37 PM To: dev@flex.apache.org Subject: Re: FalconJX: bugs in external dependencies (Google Closure Compiler), what to do? Tigran, Both Alex and I are on a Mac, so that is probably why it escaped notice until now :-) Also, the "publisher" side of the code is very much a work in progress, which deserves all the attention you can give it. I'll answer your other email some time tonight, jetlag permitting. On this one: might manipulation the path(s) before they are fed into the 'depswriter' - working from memory, might be missing the point - help avoid this low level problem? EdB On Mon, Apr 8, 2013 at 6:27 AM, Tigran Najaryan <tig...@gmail.com> wrote: > I debugged the problem that I reported earlier (generating incorrect > paths in deps.js) and looks like it is due to a bug in > com.google.javascript.jscomp.deps.PathUtil.isAbsolute. It is > implemented by simply checking that the path starts by a forward slash > which of course is incorrect under Windows. > > So what is the procedure when the bug is in an external dependency? It > is not as if it is a single call that can be avoided. > PathUtil.isAbsolute is called a few levels below in the call stack > from the > com.google.javascript.jscomp.deps.DepsGenerator.computeDependencyCalls > , which is supposed to compute and return the correct text for > dependencies file. > > I will debug this further but just wanted to understand if I should I > try to work around the external bug when I meet one or I need to try > something else? > > Tigran. > > P.S. Did anyone actually try running FalconJX on Windows or everybody > is developing on a Mac? > > -- Ix Multimedia Software Jan Luykenstraat 27 3521 VB Utrecht T. 06-51952295 I. www.ixsoftware.nl
