On Mon, 2019-05-13 at 16:14 +0100, Daniel P. Berrangé wrote: > On Mon, May 13, 2019 at 02:55:14PM +0200, Andrea Bolognani wrote: > > +# List of example programs. We need to list them here instead of using > > +# $(noinst_PROGRAMS) directly because we want to have access to the > > +# unmodified list during (un)installation, but at the same time automake > > +# might tweak $(noinst_PROGRAMS) to eg. automatically add the .exe file > > +# extension when targeting Windows. > > +EXAMPLES = \ > > + admin/client_close \ > > + admin/client_info \ > > + admin/client_limits \ > > + admin/list_clients \ > > + admin/list_servers \ > > + admin/logging \ > > + admin/threadpool_params \ > > + dominfo/info1 \ > > + dommigrate/dommigrate \ > > + domsuspend/suspend \ > > + domtop/domtop \ > > + hellolibvirt/hellolibvirt \ > > + object-events/event-test \ > > + openauth/openauth \ > > + rename/rename \ > > + $(NULL) > > + > > +noinst_PROGRAMS = \ > > + $(EXAMPLES) \ > > + $(NULL) > > I wonder if we should just directly use the _SOURCES vars instead of making > the assumption that binary names match source files. eg > > EXAMPLE_SRCS = \ > $(dominfo_info1_SOURCES) \ > $(dommigrate_dommigrate_SOURCES) \ > ..... That would (probably, I haven't actually tested it) work, however it seems to me like it would be much more likely that such a solution would eventually fall out of sync than the one I'm proposing, since when adding a new example you'd only notice the missing file if you actively performed an installation and checked the contents of the documentation directory, rather than not seeing the binary you're working on which is a much more obvious failure.

On Tue, 2019-05-14 at 10:16 +0100, Daniel P. Berrangé wrote: > On Tue, May 14, 2019 at 09:21:12AM +0200, Andrea Bolognani wrote: > > On Mon, 2019-05-13 at 16:14 +0100, Daniel P. Berrangé wrote: > > > I wonder if we should just directly use the _SOURCES vars instead of making > > > the assumption that binary names match source files. eg > > > > > > EXAMPLE_SRCS = \ > > > $(dominfo_info1_SOURCES) \ > > > $(dommigrate_dommigrate_SOURCES) \ > > > ..... > > > > That would (probably, I haven't actually tested it) work, however it > > seems to me like it would be much more likely that such a solution > > would eventually fall out of sync than the one I'm proposing, since > > when adding a new example you'd only notice the missing file if you > > actively performed an installation and checked the contents of the > > documentation directory, rather than not seeing the binary you're > > working on which is a much more obvious failure. > > I don't think such a bug is any more likely here than anywhere else > in our makefiles & we cope fine in general. We need a list of source > files & we have variables defining the source files, so it just feels > wrong to instead use a list of binary names and infer the source files > from that. As I tried to explain above, the difference is that if you're adding a new example and you forget to update $(noinst_PROGRAMS) or add the corresponding $(whatever_SOURCES), you'll immediately figure out that something is wrong because your example program will either fail to compile or just not show up. On the other hand, if you fail to update $(EXAMPLE_SRCS), nothing obviously wrong will happen, your example program will compile and run just fine, but the source code for it will not be installed as documentation. Given that we have failed to ship and install Devhelp-compatible documentation correctly for literally *years* (since 2014 AFAICT) without anyone noticing, I have little confidence we won't break it again in the future, and when that happens I'd prefer if it didn't just stay silently broken for ages. If your approach is the only one you consider acceptable, though, I'll cave in and adopt it: better broken, possibly, in the future than broken, for sure, right now :)