[Xen-users] A tale of two (build) environments: Failure to Launch

[Richard Bowser <bowser.richard@xxxxxxxxx>]

Hi everybody.  I'm becoming less of a newbie here, but I'm still stuck attempting to build xen-3.4.0 from the source.  Originally, I was stuck  by build failures complaining "gpg: Can't check signature: public key not found".  However, before I was stuck there, I started trying to 'triangulate' the source of my error by building one package under my personal Ubuntu 9.04 system, and yet another one under the Fedora Core 10 machine I'm using at school.  When both machines produced identical failure complaints, I decided I had some systematic problem, so I searched this site's archives.  I found the answer: "gpg --keyserver wwwkeys.pgp.net --recv-keys 0x517D0F0E".  Suddenly, I was past the first roadblock.  However I immediately started running into subsequent roadblocks,  only now the failure complaints were different - under Ubuntu versus under Fedora Core 10.  However there is one quite strong commonalty: "make world" fails.

Under Ubuntu 9.04: I get the most cryptic failure complaint
"mv: cannot move `linux-2.6.18/..' to `../..': Device or resource busy".  (WHAT 'device or resource'?)
Does anyone know which script is trying to move linux-2.6.18, or where it wants to move it?

I suspect that my working build directory for xen-3.4.0 just isn't in the right place.
(Does anyone know whether my working build directory needs to be in some particular location?)

It would help a LOT if I knew what the failure actually was.  Possibly I have a problem with permissions.
I've spent the last week trying to locate some script that might have its hands dirty, but no luck so far!

I KNOW that only in my target xen-3.4.0 directory, can a reference to `linux-2.6.18/..' make any sense.  So that's some help.

However, I have not the faintest clue of what `../..' in that error message actually means.  So far, I haven't been able to find documentation of "make world" process internals for building xen - anywhere.

Under Fedora Core 10: my build failures feature undefined variables.  (Maybe I am missing some needed headers.)

Here are the first few FC10 errors:

scripts/kconfig/lex.zconf.c:1623: warning: ‘input’ defined but not used

/home/rich/xen-3.4.0/linux-2.6.18/arch/i386/defconfig:48:warning: trying to assign nonexistent symbol CC_ALIGN_FUNCTIONS
/home/rich/xen-3.4.0/linux-2.6.18/arch/i386/defconfig:49:warning: trying to assign nonexistent symbol CC_ALIGN_LABELS

[long list of "nonexistent symbols" truncated]

later:
select-linux-arch: i386
/home/rich/xen-3.4.0/linux-2.6.18/scripts/mod/sumversion.c: In function ‘get_src_version’:
/home/rich/xen-3.4.0/linux-2.6.18/scripts/mod/sumversion.c:384: error: ‘PATH_MAX’ undeclared (first use in this function)

/home/rich/xen-3.4.0/linux-2.6.18/scripts/mod/sumversion.c:384: error: (Each undeclared identifier is reported only once
/home/rich/xen-3.4.0/linux-2.6.18/scripts/mod/sumversion.c:384: error: for each function it appears in.)

/home/rich/xen-3.4.0/linux-2.6.18/scripts/mod/sumversion.c:384: warning: unused variable ‘filelist’
make[8]: *** [scripts/mod/sumversion.o] Error 1
make[7]: *** [scripts/mod] Error 2
make[6]: *** [scripts] Error 2

make[5]: *** [modules] Error 2
make[4]: *** [modules] Error 2
make[3]: *** [build] Error 1
make[2]: *** [linux-2.6-xen-intree-install] Error 2
make[1]: *** [install-kernels] Error 1
make: *** [world] Error 2

Does anyone know where "PATH_MAX" is defined?

I've been studying Richard M. Stallman's GNU Make version 3.81 for guidance, but it may take a while for me to assimilate the material there and put it together with my practical observations of "make world" operational details within xen.  (Heck; and all I REALLY want to do is fix those silly errors!  Is that unreasonable?)

Anyhow folks, thanks for reading through all this!

-Rich B.
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users