Re: GNU make, gcc, and auto-dependencies

[Travis Spencer]

On Thu, Jan 06, 2005 at 10:06:34AM +0100, address@hidden wrote:
> Is it maybe because you've listed a dir as prerequisite here?
> 

That wasn't it.  I issued a make veryclean followed by a make command.
It failed though because the build directory was deleted by the
veryclean target.  I recreated the directory manually, issued the make
again, and it completed successfully.  Then, I appended an empty
string into one of my C files called kill.c, and remade it.  I got the
same error again:

pc    -c -o build/kill.o build/kill.p
make: pc: Command not found
make: *** [build/kill.o] Error 127

I think this might be because the P file is older than the object
file.  When a source file is updated, make finds out which
dependencies it has.  It concluded that the corresponding P and O
files are needed, so it tries to rebuild them starting with the
prerequisite file; however, there is no rule for building P files, so
it gives up saying that it can't find "pc," a command that make has
presumed is responsible for building P files from C files.

It is these lines in the debug output that make me think this is the case:

    Finished prerequisites of target file `build/kill.o'.
    Prerequisite        `build/kill.p' is older  than target `build/kill.o'.
    Prerequisite `kill.c' is newer than target `build/kill.o'.
    Prerequisite `vers.h' is older than target `build/kill.o'.
    Must remake target `build/kill.o'.

Why is the P file being considered a prerequisite for the O file?
Could it be because make is looking for all files with the same root
name as it seems to be saying in this line of debug output:

    Trying pattern rule with stem `kill'.
    ...
    Trying implicit prerequisite    `build/kill.p'.
    Found an implicit rule for `build/kill.o'.

I appreciate the feedback and look forward to any other suggestions.

-- 

Regards,

Travis Spencer