build: silence false errors during make clean #4553

Fixes #4548.

Automatic sanity-testing: PASSED, see http://jenkins.bluematt.me/pull-tester/p4553_1e72d5c03335e472cf5b9207419d6125fed9aec4/ for binaries and test log. This test script verifies pulls every time they are updated. It, however, dies sometimes and fails to test properly. If you are waiting on a test, please check timestamps to verify that the test.log is moving at http://jenkins.bluematt.me/pull-tester/current/ Contact BlueMatt on freenode if something looks broken.

I'd have preferred the other solution: merging in secp256k1 :-) But this is ACK too.

Well, I considered saying "just ignore this for now". Problem is that even after libsecp is merged, the makefile will only exist if it's enabled (if libsecp's configure is run). That could be avoided by always running its configure even if disabled, which is something that should probably be considered post-merge.

General principle: If you like a piece of code, you should build it. Code that is included in the default build is better maintained. Code that is conditionally or externally included rots.

Ideally we should be building libsecp256k1 every time we build bitcoin. Thus, Makefile would always exist.

Then, one could use bitcoin Makefile configury to select which library and glue gets linked at link time.

I would agree with that, and would be fine doing it that way. My main concern is initial teething pains. After merge, there are bound to be small build bugs in libsecp that cause build failures. I'd hate to see bitcoin fail to build because of an unwanted library's failure. But on the flip-side, that's the quickest way to flesh out those bugs and get them fixed, so I wouldn't complain either way.