Boulder: Makefile install target fails due to incorrect path

[sheurich]

An invalid install path is present because both CMD_BASENAMES and OBJDIR include "bin" in the path components.

https://github.com/letsencrypt/boulder/blob/9003dd4522ba08d07bfe1ec6204e42aa6252ff30/Makefile#L4
https://github.com/letsencrypt/boulder/blob/9003dd4522ba08d07bfe1ec6204e42aa6252ff30/Makefile#L14-L15
https://github.com/letsencrypt/boulder/blob/9003dd4522ba08d07bfe1ec6204e42aa6252ff30/Makefile#L54

E.g.

install: cannot stat '/go/src/github.com/letsencrypt/boulder/bin/bin/boulder-va': No such file or directory

[cpu]

Hi @sheurich :wave:

Thanks for the bug report. I think you've found some bitrot that we should clean up.

Only the rpm target of that makefile is used day-to-day by the Boulder devs or the Let's Encrypt operations team. I think the best course of action is to delete the makefile and leave a small script that can generate the rpm with fpm. We don't need the complexity of make for one glorified alias :-)

[sheurich]

Thanks @cpu 👍

It looks as though the Makefile is used by way of startservers.start():
https://github.com/letsencrypt/boulder/blob/2956b0c938c62331e25e68f73f7c6929a72884e6/test/startservers.py#L60-L69
https://github.com/letsencrypt/boulder/blob/2956b0c938c62331e25e68f73f7c6929a72884e6/test/startservers.py#L20-L24

[cpu]

It looks as though the Makefile is used by way of startservers.start():

@sheurich The plot thickens! I think these usages could probably be replaced with a vanilla go install - I'll update #3494.

I'm in favour of closing this ticket and keeping #3494 open, how does that sound to you? Is the Makefile as it exists now important to your workflow?

[sheurich]

Sounds like a good plan. I'm not using the Makefile, just looking to understand and potentially streamline the build process.

[cpu]

@sheurich Great! Keep the issues coming if you shake out more rot :-)