Nelson Elhage

It's totally possible. S3 is already mostly hidden behind the [store.Store](https://github.com/nelhage/llama/blob/main/store/store.go#L32) interface, but you'd need to add a few shims/interfaces in other places. The actual surface area for interacting with...

Yeah, I figured you'd get around to it, but it was easy and I wanted them for TakticianBot analysis. Thanks for the response! On Wed, May 4, 2016, 17:03 Chaitanya...

`llama` supports `-mem-profile`, so you should be able to measure this with something like: ``` llama daemon -shutdown; llama -mem-profile memprof.out daemon -start & # do your build here llama...

Ah, hm. I see the problem. We do parse the compile flags to extract the include path and all headers, but we currently just send files to the cloud, and...

Hm, this is breaking remote preprocessing because including all the arguments when detecting dependencies is problematic. e.g. it's very important that we strip any existing `-M*` arguments when we construct...

(When I get more time and energy to focus on Llama, it seems pretty clear to me that the next project is an integration test setup that runs the runtime...

Yeah, I think `-frewrite-includes` would work for us, but I'm not sure if there's an easy way to autodetect which one to use. Could add another configuration flag, I guess...

@http301 looks like this broke the tests – can you resolve that somehow, and also add a test demonstrating the bug that this fixes? Thanks!

The provided test case doesn't fail if I revert the rest of your changes, so if there is a real bug that's being fixed here, your test case isn't demonstrating...

I would like to see a test that demonstrates a correct behavior, that fails before the patch, and not after. The test that's currently there fails before and passes after,...