rebar3 - slow in Analyzing applications...

[Marc Worrell]

Hi,

I have (quite suddenly actually) a build problem with rebar3.
It is the newest version.

The "Analyzing applications…” phase seems to take forever.

Is there a method to see what it is doing and what might be the cause of this?

DEBUG=1 doesn’t help.

Cheers,

Marc

[Fred Hebert]

This is a full graph analysis of all files to figure out what depends on what. The first run should be slower than before, but the follow-up runs much faster as we'll only rebuild the files that have changed (along with the files that depend on the files that have changed).

A lot of the work in the application analysis was done to initially speed up large builds in companies like WhatsApp while fixing correctness at no major cost for smaller projects.

That being said, if you have a very large project with non-standard setups, you should probably come talk to us about that, it's hard to know what could be to blame there, and it's possible some options tweaking could solve a bunch of stuff, especially when you have older carryovers from rebar 2.x

[Eric Pailleau]

Hi marc

Using strace may help sometime.

Envoyé depuis mon mobile

---- Marc Worrell a écrit ----

[Marc Worrell]

Hi,

Thanks for the answers.

In the end I have removed the _build directory.
After this the first compile was indeed slow, and the next ones were ok.
Apparently something was corrupted.

Cheers, Marc

[Fred Hebert]

Interesting. I've just had a similar report from someone who saw a similar issue, and managed to save the DAG.

I'm unsure what could get corrupted and how in there since things ought to be quite repeatable, but we'll have to wait for more analyzable data to pin things properly.

[Marc Worrell]

Next time I will save the DAG and send it to you.

Thinking of it, there was one moment two rebar3 compile actions were running in parallel. This might be the cause of the corrupted file.

Thanks for your help!

Cheers, Marc

Sent from my iPad

On 17 Apr 2021, at 16:58, Fred Hebert <mono...@ferd.ca> wrote: