Martin Sebor wrote:
Andrew Black wrote:
Greetings Faird
The reason the ICC builds on windows aren't showing up is part
exporter bug, part infrastructure bug. The exporter bug is the same
bug that's been plaguing all our catastrophic failures, where runs
with missing attachments aren't exported.
FWIW, I doubt Farid knows about this or the distinction between
the exporter and the rest of our infrastructure. I myself have
a hard time keeping up with what's what and I'm pretty closely
involved in these types of things. I think I know what the
exporter is (the script that publishes the Rogue Wave nightly
test results on the Apache site, right?) but I'm not sure what
bug has been plaguing our catastrophic failures, or even that
one has been.
Correct. The exporter is what takes our nightly results and transfers
them out to the Apache site. The bug I'm referring to is the one where
results aren't included if they are missing one or more files (the build
log, the config.log file, or the config.h file). I've been working on
resolving this issue today, and should have something that can be
deployed on Monday.
Incidentally, to make sure I'm tracking the rest of what you're
saying (I thought I was at first but I'm not sure anymore), which
infrastructure are you referring to? The stdcxx Windows build
infrastructure or the Rogue Wave build and test infrastructure
that drives it? Try to keep your answer in layman terms :)
I'm primarily referring to the stdcxx windows build infrastructure,
though the problem occurs at one of the points where the Rogue Wave glue
script interface with the (generated) windows build infrastructure.
The glue scripts rely on the generated build_<compiler>.bat scripts.
These scripts are generated for the MSVC compiler, but not for the ICC
compiler. Also note that the last few lines of output from the call
to generate.bat are as follows:
Converting solution icc-9.1_ex.sln to ICC.
D:\bman5\builds\33651394\source-buildspace\etc\config\windows\generate.wsf(90,
6) (null): The handle is invalid.
So what does this mean? That there's a problem in the scripts
generated by the stdcxx Windows scripts for Intel C++ or in
the Rogue Wave infrastructure?
There appears to be at least one issue in the stdcxx Windows scripts for
Intel C++. The first issue is the condition that leads to the error
message mentioned. The second issue, which may or may not be related,
is the non-generation of the build_icc-9.1.bat file, used by the nightly
testing glue scripts.
--Andrew Black
Martin
