Toon Knapen wrote:
Rene Rivera wrote:
Very strange given that Boost.Build
explicitly disables ranlib when running in IRIX. Could you send me the
output of "bjam -v", specifically I would like to know what the
"OS=<os name>" part says.
I also tried with bjam version 3.1.4(which I use for running the
regressions tests on irix every day) and 3.1.9(fresh from the trunk)
and both seem to do the right thing.
Rene's concern about OS finally rang a bell. We have an environment
variable OS = IRIX2 for our Irix machines. That apparently overrides
bjam's variable. When I run 'bjam "-sTOOLS=mipspro" "-sOS=IRIX" ...',
it does the right thing.
I recall having this brief discussions a few years ago, when I first
attempted to use bjam to build the libraries. At the time, I gave up,
because Regex wasn't ready for bjam. But now it's worth another try.
If bjam allows environment variables to set any of its own variables,
they should be named in a way that minimizes the chance of accidental
collisions. For example, I would use something like BJAM_OS instead of
OS. You could also ignore environment variables, only allowing user
overrides on the command line. That's up to you (Bjam developers), but
the current situation is perilous.
This brings me to two further questions.
(1) How should OS be set for the other toolsets? Specifically, at the
moment I am building for darwin, gcc, intel-linux, mipspro, and
tru64cxx65. The gcc toolset would apply to various OS's (Irix, Linux
on x86, 64-bit Linux on Opteron, 64-bit Linux on Itanium). Or should I
just 'unsetenv OS' and expect bjam to do the right thing? (I would
prefer to override OS on the command line, because it makes my recipe
for running bjam easier to follow.)
(2) Can someone give me a list of variables that I need to watch out
for, in case I have set something else in my environment that will
affect bjam?
--
Dick Hadsell 914-259-6320 Fax: 914-259-6499
Reply-to: hadsell@blueskystudios.com
Blue Sky Studios http://www.blueskystudios.com
44 South Broadway, White Plains, NY 10601