RFR: JDK-8035725 Must keep microsoft VS_PATH on PATH after toolchain detection
Kumar Srinivasan
kumar.x.srinivasan at oracle.com
Tue Feb 25 14:11:06 UTC 2014
Hello Magnus,
On our team CI server we are seeing these failures on Windows,
are they related to this fix ? Or is this something else ?
Message snippet inlined below.
Thanks
Kumar
checking Checking for install src... found
checking for ant... /devtools/ant-1.7.1/bin/ant
checking for JUnit... no, deploy tests cannot be run
checking for mozilla headers... no, needed for deploy, set --with-mozilla-headers=/path/to/mozilla-headers
checking for wix... no, needed for installer, set --with-wix=/path/to/wix
checking for wscript... /cygdrive/c/Windows/system32/wscript
checking for cscript... /cygdrive/c/Windows/system32/cscript
checking for msicert... no
cygpath: can't convert empty path
cygpath: can't convert empty path
configure: The path of MSICERT, which resolves as "", is not found.
configure: error: Cannot locate the the path of MSICERT
configure exiting with result code 1
Archiving artifacts
On 2/25/2014 3:32 AM, Magnus Ihse Bursie wrote:
> After the toolchain rewrite, we could fail in various ways on Windows
> builds, e.g. like this:
> configure: Rewriting OBJDUMP to "/usr/bin/objdump"
> configure: error: Target CPU mismatch. We are building for x86 but CL
> is for "directory"; expected "80x86".
> configure exiting with result code 1
>
> The reason for this is that we save the PATH, set it to the Visual
> Studio special PATH (and other special PATH tricks on other
> platforms), do the toolchain checking, and then restore the PATH.
>
> However, on Windows, the compiler (for x86 at least) needs to have
> that special PATH at all times, since it is used for dll resolution.
> So without that path, subsequent calls to the compiler will fail.
>
> This fix makes sure we set up the Visual Studio path "permanently" for
> the rest of the configure run, so that we can always run the compiler
> after it has been detected.
>
> This was the behavior before the toolchain cleanup. And, if you're
> curious, we do setup the proper PATH in spec.gmk so that we can run
> the compiler correctly when building. :-)
>
> Bug: https://bugs.openjdk.java.net/browse/JDK-8035725
> WebRev:
> http://cr.openjdk.java.net/~ihse/JDK-8035725-keep-VS_PATH-after-toolchain-detection/webrev.01
>
More information about the build-dev
mailing list