The compile step fails on "python-gdb.py" on some buildbots.


# #### sparc solaris 10 gcc ####

running build_scripts
[51847 refs]
./install-sh -c  python-gdb.py
./install-sh: python-gdb.py does not exist
*** Error code 1
make: Fatal error: Command failed for target `python-gdb.py'


# #### x86 FreeBSD ####

running build_scripts
[53576 refs]
/usr/bin/install -c  python-gdb.py
usage: install [-bCcpSsv] [-B suffix] [-f flags] [-g group] [-m mode]
               [-o owner] file1 file2
       install [-bCcpSsv] [-B suffix] [-f flags] [-g group] [-m mode]
               [-o owner] file1 ... fileN directory
       install -d [-v] [-g group] [-m mode] [-o owner] directory ...
*** Error code 64

David, can you please take a look?

Sorry about this.

I believe this is the expansion of these fragments from the Makefile.pre.in  (indenting for clarity):

    gdbhooks: $(BUILDPYTHON)-gdb.py

    $(BUILDPYTHON)-gdb.py: Tools/gdb/libpython.py
            $(INSTALL_SCRIPT) $< $(BUILDPYTHON)-gdb.py

It looks like the second of these is being invoked in both cases with "$<" as the empty string, rather than "Tools/gdb/libpython.py" (and each with their own expansion of "INSTALL_SCRIPT").

I'm guessing that the "$<" is a GNU Make-ism that isn't available on all implementations of "make".  If so, the quick fix is probably to replace "$<" with "Tools/gdb/libpython.py" in Makefile.pre.in

I'm attaching a patch which introduces a variable for the path and uses this, rather than "$<".

I only have access to Linux machines; I've tested it on one, and it works.

I don't have commit rights, so I can't fix this directly myself.

$ make -v
GNU Make 3.81

From "info make":
`$<'
     The name of the first prerequisite.  If the target got its
     commands from an implicit rule, this will be the first
     prerequisite added by the implicit rule (*note Implicit Rules::).


(In my defence, "$<" appeared to be used already in the "Some make's put the object file in the current directory" rule).


Sorry again from breaking the build on those configurations.

Hmm.  The patch didn't apply to my current trunk checkout.  I'll look into why later.

I believe $(INSTALL_SCRIPT) was changed to $(INSTALL_DATA) in r79716 (see http://bugs.python.org/issue8032#msg102288 )

Attaching refreshed version of the patch.

Committed in r79778.  We'll see how the buildbots fare.

The buildbots seem happy.