On Mon, 28 Oct 2013 18:00:26 -0700 Gregg Smith <g...@gknw.net> wrote:
> On 10/28/2013 2:37 PM, William A. Rowe Jr. wrote: > > On Mon, 28 Oct 2013 16:58:45 -0400 > > Jeff Trawick<traw...@gmail.com> wrote: > > > >> On Mon, Oct 28, 2013 at 4:54 PM, William A. Rowe Jr. > >> <wr...@rowe-clan.net>wrote: > >> > >>> On Mon, 28 Oct 2013 13:01:09 -0400 > >>> Jeff Trawick<traw...@gmail.com> wrote: > >>> > >>>> On Mon, Oct 28, 2013 at 12:50 PM, Gregg Smith<g...@gknw.net> > >>>> wrote: > >>>> > >>>>> Just a note, > >>>>> > >>>>> > >>>>> On 10/19/2013 10:32 AM, Gregg Smith wrote: > >>>>> > >>>>>> On 10/19/2013 7:26 AM, Jeff Trawick wrote: > >>>>>> > >>>>>>> On Thu, Oct 17, 2013 at 11:08 PM, Gregg Smith<g...@gknw.net > >>>>>>> <mailto: g...@gknw.net>> wrote: > >>>>>>> > >>>>>>> I'd like to first rid the 1.5 traditional Windows build > >>>>>>> of the Release9x& Debug9x targets. Anyone against this? > >>>>>>> > >>>>>>> > >>>>>>> no concerns here; is there code that can get deleted too? > >>>>>>> > >>>>>> probably, I would assume so, I haven't ran it down that far > >>>>>> yet. > >>>>>> > >>>>> I did not do this after all nor do I think I am brave enough to. > >>>>> > >>>>> apr_escape, testescape and gen_test_char were added into the > >>>>> build with r1534053. > >>>>> .mak/.dep files were added in r1534516. > >>>>> > >>>>> We should be ready to go on the Windows side now. > >>> So, given that 1.5 remains compatible with 1.4... but projects > >>> which build APR themselves will have to adapt to the new > >>> gen_test_char or we need to make the suggested change, let's just > >>> take the httpd example. > >>> > >>> Does it make more sense for httpd 2.2 Makefile.win to detect the > >>> presence of gen_test_char and build it when encountered, or does > >>> it make more sense to compile-link-invoke gen_test_char.c? I'm > >>> happy to make the respective change later tonight or tomorrow, > >>> based on consensus. > >>> > >>> > >> Is the Windows build interface of apr 1.5.x different such that > >> httpd's Makefile.win would care? > >> > >> httpd itself doesn't use apr escape and shouldn't care that apr > >> happens to have a build utility of the same name as one of its own > >> (or something is borked). > > The httpd 2.2 Makefile.win build invokes the pcre, expat, apr, > > apr-iconv and apr-util builds project-by-project, it doesn't use > > any corresponding top-level build mechanics. This was one aspect > > that was greatly improved by moving to the httpd-2.4 model. > > > > If httpd doesn't pre-build gen_test_char before apr.mak, then the > > apr build will fail, unless we nest the new gen_test_char pre-build. > > > > I think the better solution for a one-source file, source generation > > app is to embed that build into the [lib]apr.dsp/mak files for now. > > But as an alternative, we can teach httpd 2.2 Makefile.win to > > anticipate this requirement if that project is present. That would > > mean that 1.5 apr wouldn't build with older httpd 2.2 releases, > > only the next release onwards. > This is done. The stumble point is that apr.h must be there to > compile gen_test_char.c and it must be run before ar_escape.c is > compiled. So after a lot of trying different ways to do this and get > everything to fire off at the right point in time, I just had to > sandwich it all during the creation of apr.h. > > This works in both apr.mak and libspr.mak as well. Suggestion - should we simply use the CROSS_COMPILE path when building for win32 .dsp/.mak files, instead?