Re: [lmi] PATCH: Upgrade xmlwrapp to 0.9.0

[Vadim Zeitlin]

On Wed, 11 Apr 2018 12:07:59 +0000 Greg Chicares <address@hidden> wrote:

GC> On 2018-04-11 10:46, Vadim Zeitlin wrote:
GC> > On Wed, 11 Apr 2018 01:10:20 +0000 Greg Chicares <address@hidden> wrote:
GC> > 
GC> > GC> On 2018-04-10 21:46, Vadim Zeitlin wrote:
GC> [...]
GC> > GC> I think I'd rather move xmlwrapp out of this makefile and get it from
GC> > GC> git (in which case we wouldn't have to ask you to make special 
releases
GC> > GC> for lmi)
GC> > 
GC> >  It's not a big deal to make a xmlwrapp release, it's a standard autotools
GC> > project and so it's just a matter of doing "make dist" (whatever people
GC> > say, autotools are fine for simple projects and give you a lot of nice
GC> > things for free). So I don't see any real need to change this, and I don't
GC> > expect many changes to xmlwrapp in the future anyhow, this library is
GC> > mostly in maintenance mode. Although it might be worth releasing a new
GC> > major version using C++11 one of these days...
GC> 
GC> I think commit 6e4db9e12670 might change your mind.

 Sorry, I'm not sure how exactly is this supposed to affect my mind. Just
to be clear, I actually intentionally didn't add the new files to the
makefile because they're not used yet and I planned to add them in the
patch changing the code to use RelaxNG schema for validating the input
files instead of the W3C schema. Why/how is this related to using xmlwrapp
from Git vs using a released version? I could see it as an argument in
favour of using autotools for building xmlwrapp, but my mind doesn't need
changing on this subject, I'm already all for it.

GC> Also see Vaclav's
GC> comments here:
GC>   https://lists.nongnu.org/archive/html/lmi/2009-04/msg00029.html
GC> following this line:
GC> | > One obstacle I foresee is that xmlwrapp and xsltwrapp both contain an
GC> 
GC> Now that comeau is history, there's less reason than ever to avoid
GC> autotools

 Again, this seems like an argument for building xmlwrapp as a normal
library instead of using lmi-specific hacks and I'm definitely in favour of
this. I was (and still am) just wondering about the necessity of switching
to using Git versions of xmlwrapp instead of using the releases.

GC> --especially now that we have that dependency already:
GC> 
GC> > GC> and maybe even build it with autotools, much as we do with wxPdfDoc.

 We actually don't have any dependencies on autotools yet because wxPdfDoc
includes all the generated files. But xmlwrapp doesn't and really
shouldn't.

GC> (I guess they aren't needed for wxPdfDoc because those tools were
GC> already run and their results included in the package we use, whereas
GC> xmlwrapp follows the autotools philosophy more closely.)

 Yes, exactly.

GC> but I don't use cygwin myself and don't know which of the numerous
GC> autoconf and automake files found here:
GC>   https://cygwin.com/packages/
GC> would need to be added.

 Just autoconf and automake, I think.
VZ