Possible memory leak while building from source

[hails]

When I try to compile/build from source, using either stackage and cabal, the process halts when it reaches the following step [25 of 32] Compiling Handler.Admin.Board ( Handler/Admin/Board.hs, .stack-work/dist/x86_64-linux/Cabal-1.22.4.0/build/Handler/Admin/Board.o ). I'm currently with 8GB of RAM, SSD, and 2GB of swap (file), and the building process consumes everything before halting (high I/O, system lagging etc).

[ahushh]

As a temporary solution you can decrease GHC optimization level to -O0 in Monaba.cabal

[hails]

Didn't work, but it seems to work when I use stackage stack build. First, I tried to compile in three differents Digital Ocean droplets (Ubuntu and Debian) and got some dependencies errors, even using stackage. Now, compiling (and hacking with a mix of cabal and stackage) on my Arch Linux setup kinda worked. Should we perhaps make Monaba "stackage compliant" and recheck the dependencies?

[ahushh]

Hmm, it works for me when I use cabal and latest packages with GHC 7.10.2 following my readme. I never used stackage before but I will surely take a look on it

[hails]

What Linux distribution are you using to build Monaba with cabal?

[ahushh]

gentoo

[hails]

That's strange, tried several times (even asked to a friend of mine to try buidling) and got almost the same errors (dependencies failing, memory leak, and so on). Maybe it's just me being unlucky (and being hated back by cabal) or doing something wrong.

[ahushh]

I got memory leaking too, but -O0 always helped. Don't know how to fix it another way yet. Just updated package list, inited sandbox and reinstalled dependencies, got no errors (except of nano-md5). What dependencies errors did you get?

[ahushh]

I updated dependencies versions according to yesod scaffold example, not sure if it helps

[hails]

Gonna test on a Digital Ocean droplet.

[hails]

Welp, this time everything went well. I guess I can close the issue now?

[ahushh]

Let it be open to remind me that memory leak isn't fixed, -O0 still isn't a good way

[hails]

Ok, if something like that happens again, I'll report it here.

[jeanlucaslima]

@ahushh the option -O0 should be on the documentation then