Re: compiling confusion (task cycles)

[Carl Sorensen]

On 1/22/10 11:45 PM, "Mark Polesky" <address@hidden> wrote:

> Carl Sorensen wrote:
>> I've never used make test-redo or make test-clean.  I'd
>> like to have some clarification on what they're used for.
> 
> I could be wrong, but it looks like `test-redo' accomplishes
> (automatically) what you're doing manually in your step "run
> the individual regression files..."  The CG says that it
> will "redo files differing from baseline".

Not quite the same.  In order to redo files differing from the baseline
you need to first have a make check complete.

When I'm working on a bug, the bug is demonstrated in a particular
regression test.  I want to test that regression test (and only that
regression test) until I have the regression solved.

It takes a long time to run the whole regression test suite; it takes almost
no time to run a single regression test.

> 
>> My cycle is to
>> 
>> [edit source files]
>> make (if needed)
> 
> You lost me there.  How will I know if `make' is needed?  I
> wonder if `test-redo' does this as well, whatever it is...

As John mentioned, you must do make if you edit a c++ file, a
header file.  If you only edit .scm or .ly files, you don't need
make.

> 
>> run the individual regression files to see that the
>>   regression is solved
>> Once I'm pretty sure it's right, I run
>>   make check
>>   to make sure I haven't broken anything.
> 
> One last question:  You use `-j' and `CPU_COUNT' for
> `make doc'.  Graham recommends using `-j' but not
> `CPU_COUNT' for `make'.  Are there any other `make' targets
> that would benefit from either or both of these?
> 
> Like:
>   make -j3 test-baseline
>   make -j3 CPU_COUNT=3 check

-j3 is the option that allows make to use multiple processors in parallel;
CPU_COUNT=3 is the option that allows lilypond-book to use multiple
processors in parallel, IIUC.

HTH,

Carl