On Thu, 19 Sep 2019 at 14:27, Eric Blake <ebl...@redhat.com> wrote: > > On 9/19/19 7:00 AM, Peter Maydell wrote: > > >> In an incremental build on Fedora 30, I'm now seeing: > >> > >> CHK version_gen.h > >> GEN docs/interop/qemu-ga.8 > >> No filename or title > >> make: *** [/home/eblake/qemu/rules.mak:394: docs/interop/qemu-ga.8] > >> Error 255 > >> > >> and suspect this patch introduced it. It may be that I just need to > >> nuke intermediate artifacts rather than an actual problem with the > >> patch, but I'd welcome help in identifying the problem. > > > > If you build with V=1 what does it say it's doing? > > make[1]: Leaving directory '/home/eblake/qemu/dtc' > perl -Ww -- /home/eblake/qemu/scripts/texi2pod.pl -I docs -I scripts -I > docs/interop -DVERSION="4.1.50" -DCONFDIR="/usr/local/etc/qemu" > scripts/texi2pod.pl docs/interop/qemu-ga.8.pod && pod2man --utf8 > --section=8 --center=" " --release=" " docs/interop/qemu-ga.8.pod > > docs/interop/qemu-ga.8 > No filename or title > make: *** [/home/eblake/qemu/rules.mak:394: docs/interop/qemu-ga.8] > Error 255
Do you have sphinx installed on this machine? I'm wondering if the problem here is that we have disabled the sphinx build runes but are then falling back to the "try to make any %.8 via the TEXI2MAN rule in rules.mak" [*]. On the other hand docs/interop/qemu-ga.8 is only put into DOCS if BUILD_DOCS is set, which should only happen if sphinx is available. [*] these rules are a bit bogus, because I think they will apply even if there's no other rule saying 'foo.8 depends on something.texi', and then as you can see in your command line we end up running texi2pod with an empty "$<" rather than passing it a texi file. thanks -- PMM
