Warn users not to use make -B in their workflows

[Clare72]

From my Slack post:
In the early days of ODK, the FlyBase repos (fbdv, fbbt, dpo, fbcv) were migrated over and releases were setup to run using a run_release.sh script which used -B. Very recently ODK developers have switched from promoting the use of -B, to assuming that it will not be used, and it is actually now harmful to be using it as it could result in a lot of unnecessary repetition - see #1285. I noticed today that this was resulting in the same dosdp command being run multiple times, slowing down the FBbt build.

[matentzn]

Thank you @Clare72!

I noticed this too @gouttegd; I somehow think though this existed before, but I cant remember. Basically any time you have something like this (I THINK that was it)

target1.txt target2.txt target3.txt:
    something

And you used -B I think the something just gets run multiple times. I might be wrong though, just dropping that thought here.

I personally would prefer if there was a way to make this more robust rather than warn.

[gouttegd]

I noticed this too @gouttegd;

Err, what is there to notice? The problem with the use of -B has already been identified months, if not years, ago.

I explained in details (with an example) what happens when you use -B here. Basically, -B instructs Make to “always make” (that’s the long name of the option: --always-make) a target without checking whether the target needs to be remade or not – including if Make already executed the very same target two nanoseconds ago and nothing has changed in the meantime. That is the normal behaviour of -B; when you use -B, you are explicitly asking Make to do just that.

The root problem here is that for years the ODK has been actively promoting the use of -B as a way to ensure that your ontology is entirely rebuilt from scratch. That is a terribly, horrendously wrong advice, but since we have given it for so long it is unsurprising that there must still be a lot of scripts out there that are calling a ODK workflow with make -B.

(In fact there is still at least one place in the ODK documentation where the use of -B is still recommended, as I realized a few days ago.)

Until we update the documentation, for people who might come across this ticket:

If you want to ”ensure that your ontology is entirely rebuilt from scratch”, the correct way to do that is to call make clean. You can do that in the same invocation as the one you use to trigger the build, as in:

sh run.sh make clean prepare_release

(or optionally, if if you have already refreshed the imports and don’t want to do that again during the release:

sh run.sh make clean prepare_release MIR=false IMP=false

[matentzn]

Err, what is there to notice? The problem with the use of -B has already been identified months, if not years, ago.

Sorry I should have been clearer. I am fully aware of our work to move away from -B (no Dementia yet).

This issue here is about two things:

1. Potentially add a warning mechanism. Catch the -B in run.sh and print a simple warning and say that "Use of -B is discouraged, use make clean instead, see here"
2. I think the dosdp pattern thing is the only case where -B wreaks this huge havoc. I wonder if it would make sense to avoid it. Obviously this is not on you - your stance is clear that no one should use -B, but I might be interested to figuring out how to make the Makefile "robust" for uses of -B - at least to get rid of these obvious issues.

[gouttegd]

Potentially add a warning mechanism. Catch the -B in run.sh and print a simple warning and say that "Use of -B is discouraged, use make clean instead, see here"

Quite skeptical of this.

First because reliably detecting the use of make -B could be trickier than you imagine. We can’ just search for make -B, the -B option could be anywhere on the command line.

Second because a typical ODK run outputs hundreds, maybe even thousands, of lines in the console. I think it’s quite likely that a warning emitted by the run.sh script (at the very beginning of the process) would never be seen by the user. (That’s a general problem in the ODK – we are lacking a reliable way of providing informations to the user.)

I might be interested to figuring out how to make the Makefile "robust" for uses of -B

Feel free, but be advised that I’ll object to any change that makes the Makefile even more difficult to read than it already is, if the only reason for the change is to allow people to keep using -B.

If you can do that while at the same time simplifying the Makefile, then great. :) If you end up making it more complex for no real benefit other than “people who are currently using -B can still do so”, then no. Absolutely not.

For what it’s worth, I feel your time might be more efficiently used by going through the various ODK repos and checking that they are not using -B.

[matentzn]

For what it’s worth, I feel your time might be more efficiently used by going through the various ODK repos and checking that they are not using -B.

Of course, I do that as I come across them.

reliably detecting the use of make -B could be trickier than you imagine.

I guess I did not know - and defer to your nose on this;

Feel free, but be advised that I’ll object to any change that makes the Makefile even more difficult to read than it already is, if the only reason for the change is to allow people to keep using -B.

Deal!