make Builds on Solaris Zone

[Stefan Bodewig <bo...@apache.org>]

Hi,

after a pretty frustrating afternoon[1] yesterday I hacked a new
optional "makeCommand" attribute into <workspace> in trunk[2].  All
<make> builder will use that in order to determine which make command
to use.  The default is "make".

It turns out that I also had to set the MAKE environment variable.
Still apr-make fails because it doesn't find "ar".  Since configure
should have sorted out this problem, I'm sure something else must be
fishy here.  Most likely the PATH setup is wrong, but then again PATH
is supposed to be the same in apr-buildconf, apr-configure and
apr-make.

Stefan

Footnotes: 

[1] Touched three source files and managed to introduce two syntactic
errors and one stupid mistake.  Two of them have become visible after
Gump had been running for an hour (for each attempt).  A compiler
would have found them before I committed them.  I'm not made for
Python.[3]

[2]  yes, it will be documented one day.

[3]  nd pointed me to pylint last night that I'll probably install and
use before I dare touching Gump's code again.


---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org

Re: make Builds on Solaris Zone

[Stefan Bodewig <bo...@apache.org>]

On Fri, 4 May 2007, Sander Temme <sa...@temme.net> wrote:
> On May 4, 2007, at 7:39 AM, Stefan Bodewig wrote:
> 
>> after a pretty frustrating afternoon[1] yesterday I hacked a new
>> optional "makeCommand" attribute into <workspace> in trunk[2].  All
>> <make> builder will use that in order to determine which make
>> command to use.  The default is "make".
>>
>> It turns out that I also had to set the MAKE environment variable.
>> Still apr-make fails because it doesn't find "ar".  Since configure
> 
> This means the APR configure task will fail in a more subtle way:
> it'll draw the wrong conclusions on system features from build
> failures in its test programs.
> 
> Off the top of my head, ar is in /usr/ccs/bin,

Yes, there it is.  Along with make, BTW, so we don't need to invoke
gmake at all.

I'll add /usr/ccb/bin to PATH and remove my gmake settings just to see
what will come out of it.

> S. (C weenie)

Thanks 8-)

Stefan (former C and current C# weenie).

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org

Re: make Builds on Solaris Zone

[Sander Temme <sa...@temme.net>]

On May 4, 2007, at 7:39 AM, Stefan Bodewig wrote:

> after a pretty frustrating afternoon[1] yesterday I hacked a new
> optional "makeCommand" attribute into <workspace> in trunk[2].  All
> <make> builder will use that in order to determine which make command
> to use.  The default is "make".
>
> It turns out that I also had to set the MAKE environment variable.
> Still apr-make fails because it doesn't find "ar".  Since configure

This means the APR configure task will fail in a more subtle way:  
it'll draw the wrong conclusions on system features from build  
failures in its test programs.

Off the top of my head, ar is in /usr/ccs/bin, which needs to be on  
your PATH. It is installed by the SUNWSprot package (again off the  
top of my head), which are the auxiliary build tools like as, ar,  
(sysv) make etc. but not the compiler. That's in SUNWSpro (Sun  
Workshop) or alternatively one can install gcc.

The APR (and httpd, and generally any autoconf-based system AFAIK)  
will create Makefiles that work with the make specified (or found)  
during ./configure.  I think it picks up the contents of the MAKE  
environment variable.

> should have sorted out this problem, I'm sure something else must be
> fishy here.  Most likely the PATH setup is wrong, but then again PATH
> is supposed to be the same in apr-buildconf, apr-configure and
> apr-make.

True, and should come from the calling shell. The apr-buildconf does  
not compile any C code (so does not trigger the issue), and  
compilation errors in apr-configure will be interpreted as  
unavailability of the tested feature.  These will not cause the task  
to fail, but will result in a severely skewed picture of the world  
for the apr-make task.

S. (C weenie)

-- 
sander@temme.net              http://www.temme.net/sander/
PGP FP: 51B4 8727 466A 0BC3 69F4  B7B8 B2BE BC40 1529 24AF