strange problem with `gv' install under Leopard

[Ryan Schmidt]

On Jul 18, 2008, at 03:46, Joerg van den Hoff wrote:

> thanks for taking this seriously (it sounds stupid, I'm
> sure...).

No no, it sounds like a bug, and I like getting bugs fixed, so let's  
keep at it until we do!


>> On Jul 17, 2008, at 11:49, Joerg van den Hoff wrote:
>>
>>> I'm currently in the process of a complete new install of macports
>>> on a power book g4 running leopard 10.5.3. so not much yet to mess
>>> things up...

[snip]

>>> port install gv

[snip]

>>> PROBLEM: the resulting "executable" `gv' in /opt/local/bin is a
>>> copy of the config file /opt/local/lib/gv/GV....
>>>
>>> question: my fault or a bug?
>>
>> /opt/local/lib/gv/GV is not a config file.
>> It *is* the gv binary.
>
> than the problem is already at _this_ point:
> in may case,
> /opt/local/lib/gv/GV is the same _text_ file as /opt/local/gv.
>
>> The makefile includes a step that deliberately copies this file to
>> /opt/local/bin/gv.


Upon rereading the Makefile I see I was mistaken earlier when I said  
that the it copies ${prefix}/lib/gv/GV to ${prefix}/bin/gv.

Is your filesystem case-sensitive or case-insensitive? I'm guessing  
the latter, which is the default on Mac OS X. I was using a case- 
insensitive filesystem too in my earlier test.

Testing on a case-sensitive filesystem now, what I've found is that  
after the build phase is complete, the src directory contains both  
"gv" and "GV". "gv" is the binary which should go in ${prefix}/bin/ 
gv, and "GV" is the configuration file which should go in ${prefix}/ 
lib/gv/GV. Of course a case-insensitive filesystem can only  
accommodate one of these files existing at a time.

So on your computer you have the problem that ${prefix}/bin/gv is a  
copy of the config file at ${prefix}/lib/gv/GV and therefore you  
don't have a proper gv binary so you can't run gv.

And on my computer I have the problem that ${prefix}/lib/gv/GV is a  
copy of the binary at ${prefix}/bin/gv so I don't have a proper GV  
config file and I don't know what the implications of that are.

For whatever reason (processor speed, disk speed, compiler version),  
it looks like the order in which things get built differs between our  
machines.

Clearly the build process needs to be changed so that it does not  
have case collisions like this. I filed a bug report with the  
developers of gv:

https://savannah.gnu.org/bugs/index.php?23896