Message6092

Author malte
Recipients florian, gnad, haz, jendrik, malte, silvan
Date 2017-01-12.15:34:21
Content
That error message is not great, but I assume our systems are already infected
by the "force static" virus to some extent due to the way we build the COIN
libraries. Does the same issue happen if we don't follow the instructions on
http://www.fast-downward.org/LPBuildInstructions but only build a dynamic
library for COIN? I assume this implies removing "--enable-static" and changing
the "LDFLAGS" line in our build instructions.

To be clear, I'm not against having something in the cmakefile to make the error
messages clearer, as long as the trade-off between complexity and utility is
good. One way to proceed is to incorporate your changes now and decide later
whether we want to simplify the cmakefile by removing the static/dynamic linking
options (which probably means no static linking in the most common setting).

What are everyone's preferences regarding supporting both kinds of builds going
forward? Perhaps this is something we should ask on downward-dev. If we want to
explicitly support both, then I think we must update the buildbot to provide both.

One more question: given that the dynamic/static choice only affects linking,
not compilation, it seems very wasteful to have two separate build targets,
which means compiling everything twice. The usual way that makefiles support
static and dynamic linking is to provide binaries for both settings (say
"downward-static" and "downward-dynamic"). Is this infeasible with cmake?
History
Date User Action Args
2017-01-12 15:34:21maltesetmessageid: <1484231661.64.0.0884657775586.issue662@unibas.ch>
2017-01-12 15:34:21maltesetrecipients: + malte, haz, jendrik, silvan, florian, gnad
2017-01-12 15:34:21maltelinkissue662 messages
2017-01-12 15:34:21maltecreate