Re: PDF is broken for @notation{} encoding

[David Kastrup]

"Phil Holmes" <address@hidden> writes:

> From: "David Kastrup" <address@hidden>
>
>> I'm still flabbergasted at the supposed faulty commit.  Here is one
>> theory I'd consider more plausible:
>>
>> commit 5eca56fae0faa2db9cf7f12903e1a06c42b2af0d
>> Author: Walter Garcia-Fontes <address@hidden>
>> Date:   Sat Feb 7 20:00:15 2015 +0100
>>
>>    Doc-ca: texinfo.tex and txi-ca from upstream to fix problem with
>> Catalan interpunct
>>
>> This commit contains the following diff in tex/texinfo.tex:
>>
>> @@ -8821,6 +8949,7 @@ directory should work if nowhere else does.}
>>       \catcode\count255=#1\relax
>>       \advance\count255 by 1
>>    \repeat
>> +
>> }
>>
>> This diff introduced a spurious empty line resulting in \par into the
>> command \setnonasciicharscatcodenonglobal which may be used in several
>> different situations, possibly reading indexes and/or macros.
>>
>> This would _totally_ believably match the reported symptoms (and likely
>> warrants fixing).  But it is definitely a different commit than
>> reported.
>
>
> I may be getting confused over the order of commits.  If I look at the
> Savannah web page
> (http://git.savannah.gnu.org/cgit/lilypond.git/log/?ofs=100), I see
> Werner's commit stem-tremolo-style.ly: Minor doc improvement
> e1081454de4c8167f3f010912de061822226f61d immediately before the update
> texinfo.tex from upstream commit: so I have been checking out the
> tremolo commit and then the texinfo commit, and the former gives no
> errors and the latter gives the extra line feeds.

Oh.  Savannah just lists all commits in the repository in chronological
order.  But those are commits to different branches.  So your points of
comparison differed by much more than the difference introduced by the
latter commit.

> However, if I look at the git history with gitk, it appears that the
> texinfo commit occurred just after 2.19.17 and the tremolo commit just
> after 2.19.19.  So it looks like savannah's order and the actual git
> history are not identical.
>
> Could you tell me how to be sure what order the commits occurred in,
> please?

There is a _chronological_ order to them, but no _topological_ one as
they occured in separate branches.  Here is what
git log --date-order --oneline --graph has to say about their relation:

*   e8d5061 Merge branch 'master' into translation
|\  
* | f11e0db Doc-ca: fix remaining errors
* | 596736f Doc-es: add requested documentencoding
* | 78a1db7 Doc-de: add language specification where needed
* | b75eb0e Doc: typo in @documentencoding
* | 9f3c771 update texinfo.tex from upstream
| * e108145 stem-tremolo-style.ly: Minor doc improvement.
| * 8dd74d9 Rename makefile variable TMP
| * 3ca8114 Updated patch for issue 155
| * fac5dc1 Issue 4351/2: Rename Translator_void_method_ptr to 
Translator::Callback
| * 20a1118 Issue 4351/1: Rename Engraver_void_function_engraver_grob_info to 
Translator::Grob_info_callback
| * 0e77b85 Release: bump VERSION.
| * df49959 Release: bump Welcome versions.
| * 5854ae4 Release: update news.
| * 5d4ef74 PO: update template.
| * ec5bbd1 Release: bump VERSION_DEVEL.
| * f758afb Remove cygwin_conv_to_posix_path
| * 3740ac2 Issue 4348: Part combiner: move direction handling out of iterator
| * 14261ea Issue 3799: New satb.ly built-in template and template framework
| * 8f2f4d6 Doc: Issue 4349: Clarify where changes to beatStructure should be 
placed
| * 72ec71b More flexibility for tremolo slashes
| * 2eb22fd Set CFLAGS and LDFLAGS to build python modules
* | 1a1f06a Doc-ca: fix non existing xref
* | c55c0ee Doc-ca: Applying the patches contributed by Jean-Charles Malahieude 
to the Catalan translation
* | 7fffe64 Doc-ca: Replaced l.l by l·l since in principle not it is save to 
use it
* | d10fd0c Doc-ca: Fixes for Catalan translation
* | 7917900 Doc-ca: Priority 4 Catalan Translation
* | 5eca56f Doc-ca: texinfo.tex and txi-ca from upstream to fix problem with 
Catalan interpunct
* | b3cfc92 Doc-ca: Translated string seen in home page of lilypond.org about 
hosting
| * e0af94b Replace C++ (in)equality checks with proper SCM syntax
| * 2373435 Issue 4343: quoteDuring music should be sent to midi file
| * 205b8c5 Issue 4333: Cue notes should not be included in ambitus
| * b2ca0fe Remove unused music type general-music
| * af1b0da Add sans-serif and monospace fonts
| * 9d84914 Part combiner: allow a2 chords
* |   f809d8d Doc-hu: Update fundamental.itely: Localize comments in examples
|\ \  
| | * 1010f57 LSR updates
| | * 3e8523a Fix TODO with markup padding and fix buggy example in LM
* | | c6ee0c8 Web-it: fix missing @divEnd
| | * cbe0ea8 Issue 3615: more consistency with key sig grobs
| | * bd68245 PO: updates Danish from FTP
| | * ae17151 Release: bump VERSION.
| | * c284000 PO: update template.

So you see that their commits (marked by * ) happen in branches (marked
by | ) that have an individual long history of their own.

> FWIW I started with git bisect but found problems with builds failing,
> so resorted to a home-grown style of bisecting - hence needing to
> establish the order of commits.

Huh.  Git bisect would have heeded the _topological_ order and would
have made it more likely you'd have found the correct commit.  There is
a git bisect command for reporting an untestable commit, namely "git
bisect skip".

So my best guess is that the change I identified is the culprit here and
that your homegrown bisection did not produce the right commit.

As you can see, the commit you identified is _chronologically_ just
above a commit in the other branch, but both branches have a long
separate history of commits by which they differ previous to that.

-- 
David Kastrup