.gpi in subdirectories

[GoogleCodeExporter]

What steps will reproduce the problem?
Building a .gpi file that resides inside a subdirectory which contains the data 
files.

What is the expected output? What do you see instead?
The make fails, because the gpi.d file doesn't contain the paths for the data 
files, just for the eps 
and gpi.d file. e.g. inside the gpi.d file on the penultimate line reads 
something like

Data/Directory/output.eps Data/Directory/output.gpi.d data1.dat data2.dat 
data3.dat

Prefixing data1.dat etc. with Data/Directory/ allows the make to finish 
successfully, as it find the 
files when it checks they exist, and doesn't look for a makefile process to 
generate them.

What version of the product are you using? On what operating system?
Makefile 2.1.34, CentOS 4.8

Original issue reported on code.google.com by amplit...@gmail.com on 8 Nov 2009 at 11:20

[GoogleCodeExporter]

Updated to 2.1.35, and despite the fact I can't see a relevant change in the 
log, it now seems to work. Not sure 
what the deal was... Sorry for the clutter!

Original comment by amplit...@gmail.com on 8 Nov 2009 at 11:26

[GoogleCodeExporter]

Interesting - there may have been a minor correction to a sed call that allowed 
directories.  I'm not sure why it suddenly works.

Incidentally, subdirectories are simply not supported, so when they work, it's 
luck more 
than intent.  The files build in subdirectories, for example, don't get cleaned 
with make 
clean.  That may actually be why it suddenly "works".

Original comment by shiblon on 8 Nov 2009 at 1:35

[GoogleCodeExporter]

Any further news on this?  Does it actually work, or was the makefile just 
ignoring it 
because it was already built by hand?

Original comment by shiblon on 19 Nov 2009 at 6:54

[GoogleCodeExporter]

sub-directories are never going to be fully supported by the makefile, since 
they strike 
at the heart of some of make's limitations, and since the output in those 
directories 
would never be cleanable.  Marking WontFix.

Original comment by shiblon on 19 Nov 2009 at 7:15

• Changed state: WontFix

[GoogleCodeExporter]

Sorry, I should have followed up. Indeed, the files were already built, which 
is why it started working.

Original comment by amplit...@gmail.com on 20 Nov 2009 at 9:57

[GoogleCodeExporter]

Yeah, I kind of figured.  I wish it were easier to do this, as it's a fairly 
obvious need.  
That said, I had very little trouble navigating a directory with hundreds of 
files in it, 
since the makefile does a pretty good job of cleaning up after itself.

Sigh.  I am thinking about how to rewrite this using something else so that 
these issues 
just go away.  "make" has some fundamental limitations that make this sort of 
thing 
difficult.

Original comment by shiblon on 23 Nov 2009 at 4:13