tmap 3.2.0

[2bits]

• Upgrade to version 3.2.0
• Use git and a tag for stable so that we pick up the submodule.
• Add options for 32-bit, Alignment, Google Perf Tools, and NoColor
• Add deps on 'automake' and 'libtool' to generate 'configure'
• Add a conditional dep on google-perftools if optioned.
• Add env :std and fails_with :clang otherwise missing symbols.

[2bits]

@jacknagel is this submodule issue is on my end or does HB have any control over it?

This formula builds tmap-3.2.0 which includes a git submodule called samtools at a specific commit. When I try to build this, it pulls samtools all right, but I don't think it pulls the right commit.

Here is the tmap code repo showing samtools submodule
Here is what I get when I try to build it

Thanks.

[adamv]

We need to kick upstreams to provide real tarballs; git-flow has submodule issues too.

[jacknagel]

It looks like it is checking out the correct submodule commit to me.

[jacknagel]

i.e. it matches the submodule commit in the superproject at the 3.2.0 tag: https://github.com/iontorrent/TMAP/tree/tmap.3.2.0/src

[2bits]

Ok thanks. I tried to use git log in the samtools directory to have it tell me what commit it was at, but all I got was the log for the superproject, and I didn't see the commit listed in the verbose output. Is there an easy way I can check what commit it's checking out when I run into this in the future?

[jacknagel]

git submodule status (run from the root of the superproject) should list the checked out commits of all submodules.

In the cache, it is slightly complicated by the fact that submodules are not even cloned until the first installation attempt, so a naive brew fetch tmap will only clone the superproject (very recent git versions can recursive into submodules during a fetch operation, but git-clone can't and we don't require the most up-to-date git anyway).

[2bits]

I sent a bug report upstream: iontorrent/TMAP#3 about the missing symbols I get atm.

[nh13]

Thanks for taking the time to update the Formula. I am seeing a similar error. When go to ~Library/Caches/Homebrew/tmap--git after the failed install, and run the same system commands (autogen.sh, configure, make) it works fine, so somehow transferring from the checkout directory to the build directory is not working. This affects sources from the submodule.

https://gist.github.com/3908018

[nh13]

More debugging, working with gcc/g++ it works, but with cc/c++ it fails, hmmn.

[2bits]

@nh13 thanks for stopping bye and mentioning that. @jacknagel Why would this build successfully inside the cache but not when it's cloned into the buildpath? I even set GIT_DIR during testing as one of the first things just in case.

[2bits]

I got it to work using std env and llvm. I have no idea why that didn't work before when I tried it, but I must have forgotten to do it. I might not even need the patch. Will test some more later tonight.

[adamv]

Patch at least needs a comment

[2bits]

Rebased and patch removed. It wasn't needed (to stop an issue I thought might be caused by -msse3).