Extremely small value in making this optional. bitcoin-cli almost always has utility.
Further, distro builders and other packagers typically build everything, then cherry-pick if something needs to be elided. Developers normally build bitcoin-cli.
The target audience for this option approaches zero. I’m sure it satisfies some engineer’s OCD somewhere, but is otherwise of little practical value. :)
Extremely small value in making this optional. Reduce complexity slightly.
Heh, I suppose that engineer would be me.
I have no case against the removal, but I can’t see any real benefit in it either.
FYI, my reasoning for always giving the option to disable (this happens quite often): I’m building a program/lib for an architecture where the binary I want builds fine, but compile borks on some other binary that the upstream devs decide is super-mega-important (it’s not). Now I have to patch my build to work-around.
Granted that’s not likely to happen here. Just explaining the logic.
ACK if others agree, but that doesn’t mean I like it :)
@theuni It’s really a question of scale. A zen balance must be achieved. :) For the GUI, with massive additional dependencies and compiled modules, it makes a lot of sense.
The default, on the other hand, should be to avoid adding options without good reason. I would rather fix a bug, normally, than add a –paper-over-bug option that is very rarely used, in highly unusual and soon fixed scenarios.
In theory, every additional option increases the test matrix. Removing options reduces complexity. In this case, admittedly, it is a very tiny marginal move.
I tend towards NACK. It’s not that I disagree with you, and were the build system to be redesigned from scratch again in maybe 5 years, I’m fine with not adding an option for this triviality.
But this is already in releases (starting from 0.9). I don’t see why I’d want to burden anyone that already uses this option with having to change their script because somehow it is not ‘pure’. Imagine being on the side of the user and being annoyed about this.
It adds trivial overhead, and insisting on removing it seems like a bit of “OCD” as well. Sorry.
Where is the evidence that anyone uses this? The feature was added speculatively “not likely to happen here” - The author. The builders of the binaries that reach the vast majority of users don’t use it.
It was a minor bit of over-engineering. Feature added before there was a need or call for it.
You should not add features that might, in theory, be used in some rare case sometime in the future by somebody.
Come on, is it really so hard to type “make bitcoind” etc. rather than just “make”?
You don’t need to build extra gadgetry to selectively build binaries.
You’re proposing a change so IMO the burden of evidence is on you. We don’t have statistics about this at all.
To be clear: If it were broken and no one noticed for a long time, you’d have a good case for removing it.
You can argue that this should not have been added, my retort then would be that you should have catched it either before it was added, or at least sooner.
Come on, this is just a few trivial lines in the build script. What are you worried about? Is this even worth the time discussing this?
By that logic we are burdened with useless crap forever…
If removing unused bits of lesser importance is so difficult, it is disappointing to consider deprecating and removing anything more ambitious. We’re talking about a feature where the -intended target audience- is a handful of knowledgable devs.
It’d be nice if “make bitcoind” worked…
Side note: Gentoo does in fact use this option. bitcoin-cli and bitcoind are separate packages.
IMHO bitcoin-cli and bitcoin-tx should be similar. Neither has additional dependencies, and both are only optionally needed. So either have an option for both, or for neither.
I lean towards having an option for both, as it’s a tiny change only, and is consistent with –with-gui which is hard to avoid with so many extra dependencies. I agree with Jeff about minimizing configuration combinations, but these are examples of building some more binaries from the same source. The ways that this can result in modified behavior are very minimal imho (as opposed to say –disable-wallet, which actually changes the code being built).
Ok, conclusion: let’s change –with[out]-cli to –with[out]-utils, and include bitcoin-tx in the same bag.
We could make --with[out]-cli a temporary alias for the other one if we really care about not breaking things.
Are you going to update this @jgarzik or should we make a new pull?