Re: build problem with lilypond-2.17.24 & 25

[David Kastrup]

David Kastrup <address@hidden> writes:

> Thomas Klausner <address@hidden> writes:
>
>>> Can you also report which of the preexisting files in lily/out/ gets
>>> overwritten by the problematic build, and which not?
>>
>> The header file is not recreated, but the cc file is:
>>
>> -rw-r--r-- 1 pbulk builder 258460 Sep 2 07:14
>> work/lilypond-2.17.25/lily/out/parser.cc
>> -rw-r--r-- 1 pbulk builder 9822 Sep 2 07:14
>> work/lilypond-2.17.25/lily/out/parser.dep
>> -rw-r--r-- 1 pbulk builder 7646 Aug 25 14:42
>> work/lilypond-2.17.25/lily/out/parser.hh
>> -rw-r--r-- 1 pbulk builder 597408 Sep 2 07:14
>> work/lilypond-2.17.25/lily/out/parser.o
>
> I took a look at the tar file.  The time stamps of parser.cc, parser.hh
> and parser.yy are identical as far as I can see which is not helpful for
> make.  Maybe we should take a look at the process creating the tar
> file.

This would appear to be relevant (particularly the last sentence):

commit c20fff645e325d392d2588721144f4204d0ef9da
Author: John Mandereau <address@hidden>
Date:   Wed Aug 1 12:22:43 2012 +0200

    Make distributed tarball from Git file list
    
    This removes the requirement of having a GNUmakefile
    in each and every directory of the source tree; this commit
    also deletes GNUmakefiles made unnecessary this way.
    
    dist-toplevel-txt-files toplevel target is also removed,
    because it is unneeded outside of dist target and it
    prevents linking of toplevel generated txt docs to
    distdir when these docs are already generated.
    
    In addition, on suggestions from Han-Wen, if dist is made with a
    source directory tracked by Git, then
    * the source directory is required to have no uncommitted changes nor
    non-ignored untracked files,
    * the time stamp of all files in the tarball is set to the time stamp
    of the head of the checked-out branch.


Ouch.

-- 
David Kastrup