On 02/11/2017 03:52 PM, Assaf Gordon wrote: >> Attempts to do 'make install-html' fail with: >> >> make[1]: Leaving directory '/usr/src/sed-4.4-1.i686/build' >> Making install-html in po >> make[1]: Entering directory '/usr/src/sed-4.4-1.i686/build/po' >> make[1]: *** No rule to make target 'install-html'. Stop. >> make[1]: Leaving directory '/usr/src/sed-4.4-1.i686/build/po' >> make: *** [Makefile:3027: install-html-recursive] Error 1
> The switch to non-recursive makefiles was modeld as closely as > possible after coreutils, and I see the same error when trying > 'make install-html' in coreutils. > > Not that this makes it OK, but it hints that this is not > specific problem in sed's build system. I worked around it while packaging sed for Cygwin by doing: --- origsrc/sed-4.4/po/Makefile.in.in 2017-02-03 19:15:46.000000000 -0600 +++ src/sed-4.4/po/Makefile.in.in 2017-02-11 09:30:51.544239400 -0600 @@ -473,3 +473,5 @@ force: # Tell versions [3.59,3.63) of GNU make not to export all variables. # Otherwise a system limit (for SysV at least) may be exceeded. .NOEXPORT: + +install-html: so maybe that argues that the bootstrap process should tweak po/Makefile.in.in and/or makevars in such a manner. I don't know if it is a weakness of gettext in general that it's makefiles do not play nicely with all of the usual supported targets. > > If there's a suggested solution, I'm happy to use it. > Otherwise, I'll try to dig in and see how to fix it, > but it might take a bit longer. Maybe one of the gettext developers has an idea, since it is the generated po/Makefile that needs the tweak? Adding them in cc. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
