[Lldb-commits] [lldb] [lldb-dap] Improving consistency of tests by removing concurrency. (PR #165496)

[David Spickett via lldb-commits]

DavidSpickett wrote:

> We should totally be able to handle an incomplete packet (and I assume/hope that's what you're referring to when you say "fix our bot").

I should have been clearer: our as in Linaro's bot.

I mean that I'm going to give lldb-arm-ubuntu fixed resources so it doesn't have so many random delays. Separate from whatever you decide to do with lldb-dap. It's something I should do anyway, we used to run a lot of bots on shared machines but over time more and more have fixed resources because of situations like this.

I think it's actually unfair to require lldb-dap tests to be stable on that bot because it can vary so much. On the AArch64 Linux and Windows bot where they have the whole machine, it's more fair to expect it. Took me a while to realise that but got there eventually :)

Trying random delays to destabalise them, yes that sounds like a good experiment to do. I have tried this with the "stress" utility, not to much success but I'm sure there are better ways.

https://github.com/llvm/llvm-project/pull/165496