1637546 - mach bootstrap (?) removes my ~/.mozbuild

Status: RESOLVED INVALID

(Firefox Build System :: Bootstrap Configuration, defect, P3)

Description

[David Teller [:Yoric] - still alive but not very active]

Every so often, my entire toolchain fails me and I realize that /.mozbuild has vanished. Whenever that happens, I waste a few hours trying to reconstruct it. Typically, in such cases, ./mach fails, apparently because of hg errors which are due to to extensions not being present, which in this case is due to the fact that they should be in /.mozbuild.

I'm fairly confident that this happens at some point during a call to ./mach and I have the impression that this happens when I call ./mach bootstrap to update my toolchain.

I'll investigate further.

My config:

• macOS 10.15.4 (Catalina) – but I encountered the same issue with previous versions of macOS ;
• Homebrew ;
• in case it matters, my source files are on a different mount as my home directory.

Comment 1

[David Teller [:Yoric] - still alive but not very active]

I should add that, once this happens, neither ./mach bootstrap nor ./mach clobber python seem to work anymore.

Comment 2

[Ricky Stewart]

This would be very concerning. When you have more information or reproduce, let us know. (As you can imagine, this is not enough information right now to approach doing a bugfix or anything.) For what it's worth, my day-to-day configuration is exactly the same as yours and I've never seen anything remotely like this; except I don't have a separate mount for my source, so that could be the confounding factor. (Oh, and I use git and not hg, but I don't see how that would be relevant.)

What's the deal with this extra mount for your source?

I should add that, once this happens, neither ./mach bootstrap nor ./mach clobber python seem to work anymore.

Can you go into more information about what "not working" means? Is there an exception, what is the exception, etc.

Comment 3

[David Teller [:Yoric] - still alive but not very active]

Well, it just happened again.

This took place just after I realized that this tree didn't rebuild C++ code, so I assumed it was configured for artifact builds and I attempted to reconfigure it for regular builds. The ./mach bootstrap failed with an error message that sounded like bug 1636797. I attach the error message, although it's probably not related to the current bug.

Here's my bash history:

./mach bootstrap # Failed.
rm `find obj-x86_64-apple-darwin19.4.0/ -name "*.pyc"` # Workaround for bug 1636797
hg commit -m "WIP: Tracing unnamed threads"
hg pull --rebase
./mach build && ./mach run
./mach boostrap # Fails again, same error message, still probably unrelated to the current bug.

I'm almost sure that I realized that ~/.mozbuild had vanished after the first call to ./mach boostrap.

Also, I found out what happens with ./mach clobber python. Apparently, I was missing purge = in my .hgrc. I assume that it should be added to ./mach bootstrap. Now that I have added it, ./mach clobber python seems to work.

edit The latter is bug 1637709.

Comment 4

[David Teller [:Yoric] - still alive but not very active]

Attachment: bootstrap.log

Comment 5

[David Teller [:Yoric] - still alive but not very active]

Ok, more investigating.

It turns out that my /.mozbuild is actually in /Volumes/Code/dteller/.mozbuild. It's a good place to be. So the description of the bug here is probably invalid. There is something that sometimes breaks all my calls to hg and some calls to mach, but it's not a missing .mozbuild.

So, I suggest we close this bug as INVALID. The missing /.mozbuild is a false lead. I'll reopen when I have more details.