| RE: [cdt-dev] PathEntryUtil should honor cygwin absolute paths |
|
Yes, please be clear that using cygwin on
Windows is optional. We have already caused a lot of confusion for poor mingw
users (and there is a bug open on that) and I will be introducing the
integration with the Windows SDK soon. On the other hand, another thing to
think about is cross-compilers that run on cygwin which use cygpath. In the
end, it really is the build tools that decide whether you are using cygwin or
not. All in all, cygwin is good and evil all at
the same time. The JRE doesn’t know about the cygwin file system. Maybe
this is a use for the Eclipse File System to allow users to use cygwin paths while
working in Eclipse, maybe not. Cygwin does kindly record it’s mappings in
the Windows registry. But whatever we do, we can’t mess up the use of
native paths on Windows, and remember ‘/usr/include’ is a valid
native path which almost always has a different meaning with cygwin. From:
cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Sennikovsky, Mikhail Hi Norbert, My point is that the path translation should be performed in the
Build System within the mechanism of path entry from option values calculation.
See the
org.eclipse.cdt.managedbuilder.internal.core.ManagedBuildInfo.getManagedBuildValues()
method for more info. When calculating the path entries, paths specified in the option
values should be translated to OS path if necessary and Path entries containing
regular OS paths should be created. The tool definition may provide the path translation mechanism by
specifying a java call-back. Build System would define a public implementation
of such a call-back that would translate cygwin paths into windows paths, so
windows tool definitions might reuse this call-back. Mikhail From:
cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Ploett, Norbert Mikhail, I
found the place in the PathEntryUitl by tracing the message in the Problems
back to it's source. I am not quite sure where to start in the MBS. I mean a
PathTranslator class could be specified per toolchain or even per tool, but who
will evaluate the information. Can you give me a pointer, where the strategic
place for a change would be? Thanks, Norbert Von:
cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] Im Auftrag von Sennikovsky, Mikhail I guess the logic of path conversion should reside in the Build
System, but not in the Path Entry mechanism, since it’s the build system
who knows what tool-chain is being used and how the tool-chain treats path
information. The Path Entry in its turn should work with OS paths as it does
now. So I propose making the enhancement in the Build System if
necessary to properly convert the include path information when calculating
path entries. Mikhail |