Concept ACK

IIUC we do still check BDB functionality in plenty of other tests, so this just loses the memory sanitizer coverage. Since we shoved the legacy (BDB) wallet into a box (LegacyScriptPubKeyMan), we probably won't be touching its internals (often). Just have to remember, when we do, to be a bit extra careful.

Are you sure this actually removes BDB from that task? E.g. BDB_PREFIX is still used in BITCOIN_CONFIG. Maybe explicitly opt-out with --without-bdb (also for clarity).

Thanks, fixed. NO_BDB=1 is already set.

Thanks, fixed. NO_BDB=1 is already set.

That's odd; then what was the point of building it? Only compile time checks for dependency itself? But not any of the call sites, and no runtime checks?

Building bdb from depends was never supported with msan, which is why it used the script. This works, as configure will pick up the non-depends bdb. However, now that the non-depends bdb is removed, there is no need to specify --without-bdb, as bdb isn't even built.

Oh wait, NO_BDB only affects the depends itself, but you can still build it from script. With other depends it doesn't work that way, because it changes the configure settings to match. Fairly confusing, hence my suggestion to make it explicit with --without-bdb.

I think it is better to keep it as is now, because:

• No bdb build is known to pass, so there should be no way to re-enable one without making CI red.
• If there was one that passes, having to modify a single place seems less confusing than having to modify two places at the same time.

Looks like this fixed itself, but we should still consider doing this anyway in the future.

Concept ACK. I had a similar change in an MSAN branch.