Re: [octave forge] (generate_html) fprinf failure during html generation

[JuanPi]

Hi Mike,

Thanks a lot for investing time on this.

The whole effort to separate matgeom form geometry was exactly to
avoid editing the upstream code, sync becomes a real issue.
I could imagine however that a hook when merging from upstream could
be used to replace all % with #, being careful not to replace inside
format strings and such.
I would rather avoid doing this if possible, but if there is no other way...

On the other hand, generate_html should be tolerant to the % symbol,
so I filed the bug.

Regarding the issue with the sentences, I will report to upstream, see
if they accept it (I do not see why they wouldn't)

Thanks!


On Tue, May 14, 2019 at 12:44 AM Mike Miller <address@hidden> wrote:
>
> On Mon, May 13, 2019 at 23:45:49 +0200, JuanPi wrote:
> > Hi all,
> > I am preparing the release of matgeom and geomtry and when I am trying
> > to generate the html for matgeom I get
> >
> > > pkg load generate_html
> > > generate_package_html ("matgeom", "OF/matgeom-html", "octave-forge");
> > error: fprintf: invalid format specified
>
> In the context of printf-style functions, the term "invalid format"
> usually means that an unescaped percent character is part of the
> template string. Typically, this means the printf function is called
> incorrectly with unsafe input, for example
>
>     printf (my_string);
>
> instead of
>
>     printf ("%s", my_string);
>
> So this is a bug in the generate_html package, when the first help
> sentence of a function contains a literal '%' character.
>
> A simple test case would be
>
>     ## -*- texinfo -*-
>     ## @deftypefn {} {} myfunc
>     ## This function is 10% ready.
>     ## @end deftypefn
>     function myfunc ()
>     endfunction
>
> I can reproduce the exact error with a trivial package containing this
> function.
>
> So actions to move forward from here
>
> 1. file or fix the bug in generate_package_html
> 2. see if it's easy to avoid using '%' in function doc strings
>
> I notice that matgeom in particular uses Matlab-style help blocks, and
> typically does not end the first line with a period. So you may find
> that none of the first lines of all doc strings have a percent, but
> Octave will read multiple lines until it sees a period. Look at
>
>     get_first_help_sentence ('circle3dPoint', 200)
>
> for an example from your package that returns much more than a summary
> sentence, and includes a '%' character that triggers this error.
>
> --
> mike



-- 
JuanPi Carbajal
https://goo.gl/ayiJzi

-----
“An article about computational result is advertising, not
scholarship. The actual scholarship is the full software environment,
code and data, that produced  the  result.”
- Buckheit and Donoho