trouble compiling Masurca on Ubuntu

[bzip2]

Hi. Are you willing to make a Masurca binary available for Ubuntu? I'm interested in using Masurca for transcript assembly as described in the StringTie paper, but I'm surrendering after spening a couple of hours trying to install it to no avail.

The first challenging dependency is gcc. I tried the Ubuntu default (7.2.0) and it complains about -V and -qversion. I commented those out but ran into the problems with boost (below). Because of the cryptic fatal error messages, I also tried installing gcc v4.7 for Ubuntu but it behaves the same way as 7.2.0 (doesn't recognize -V and -qversion). There doesn't seem to be an Ubuntu option at CERN.

The second challenging dependency is boost. FYI, it is not mentioned here. I installed boost and tried pointing install.h to that folder, but it reports that it can't find a working installation of boost:

configure: Detected BOOST_ROOT; continuing with --with-boost=/usr/include/boost/
checking for Boost headers version >= 1.46.0... no
configure: cannot find Boost headers version >= 1.46.0
## ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ ##
## Could not find a working installation of Boost. Set BOOST_ROOT to the path where the Boost headers are installed or set BOOST_ROOT=install to have it downloaded from the Internet and installed locally. For example: BOOST_ROOT=install ./install.sh ##

The same thing happens if I modify install.sh and pass that directory directly to configure (./configure ... --with-boost=...).

I copied the boost directory into if global-1 and got a different error:

checking for Boost headers version >= 1.46.0... /usr/include
checking for Boost's header version... 1_58
checking boost/icl/interval_set.hpp usability... no
checking boost/icl/interval_set.hpp presence... yes
configure: WARNING: boost/icl/interval_set.hpp: present but cannot be compiled
configure: WARNING: boost/icl/interval_set.hpp:     check for missing prerequisite headers?
configure: WARNING: boost/icl/interval_set.hpp: see the Autoconf documentation
configure: WARNING: boost/icl/interval_set.hpp:     section "Present But Cannot Be Compiled"
configure: WARNING: boost/icl/interval_set.hpp: proceeding with the compiler's result
configure: WARNING:     ## ------------------------------- ##
configure: WARNING:     ## Report this to gmarcais@umd.edu ##
configure: WARNING:     ## ------------------------------- ##
checking for boost/icl/interval_set.hpp... no
configure: error: cannot find boost/icl/interval_set.hpp

Thanks.

bz

[gmarcais]

Hi bz,

you do not need to modify install.sh. Actually, the error message says it all: you need to set the environment variable BOOST_ROOT and NOT pass the switch --with-boost. The easiest is to have install.sh install a version of Boost locally. Run the following command and you should be all set:

BOOST_ROOT=install ./install.sh

[bzip2]

Thanks - setting BOOST_ROOT=install worked when I started from scratch.

The build now fails at a different point:

checking for zlib.h... no
...
configure: error: in `masurca/MaSuRCA-3.2.6/global-1':
configure: error: Cannot find zlib.h header. Install zlib development files (e.g. zlib1g-dev)
See `config.log' for more details

Config.log says this:

configure:16992: checking zlib.h usability
configure:16992: g++ -c -DNDEBUG -std=c++0x -g -O2  conftest.cpp >&5
conftest.cpp:68:10: fatal error: zlib.h: No such file or directory
 #include <zlib.h>
          ^~~~~~~~
compilation terminated.

I have zlib1g-dev installed. I tried reinstalling it for good measure, but it's already at the latest version. I tried setting CPP_INCLUDE_PATH to the directory where zlib.h is installed. I also tried copying zlib.h to global-1 and other subdirectories, but I keep getting the same error. Is there an undocumented environment variable or compilation flag that needs to be set?

[gmarcais]

bz, I don't quite know what is going on with your system, but it should not give you that much of a hard time. MaSuRCA is compile on Ubuntu regularly.

First, make sure that you have the build tools installed on your system: sudo apt-get install build-essential. Second, please start from a freshly downloaded distribution of MaSuRCA by downloading this tar ball: MaSuRCA-3.2.4.tar.gz.

Then run:

tar zxf MaSuRCA-3.2.4.tar.gz
cd MaSuRCA-3.2.4
BOOST_ROOT=install ./install.sh

And that should be it. Do not modify the install.sh or configure scripts. That is unlikely to get you anywhere. Also, not all gcc errors that you read in the config.log files are actual errors. Many of them are just tests to discover which compiler or where libraries are installed, and it is OK for them to fail. You should only read config.log if the configuration failed.

Now, regarding you zlib issue. If you have zlib1g-dev installed, then the command dpkg -L zlib1g-dev | grep zlib.h should return, among other things, /usr/include/zlib.h. If not, install it with sudo apt-get install zlib1g-dev.

You do not need to set any variable or do anything particular beyond installing zlib1g-dev for it to work.

[bzip2]

Hi again. The problem turned out to be that I also had g++ installed with conda, so /usr/include was not in the default include path. Once I used the default g++, everything compiled fine. Thanks for your assistance.