Bug #473
BuildGenerator 0.6.10 fails to resolve boost
| Status: | Closed | Start date: | 2016-03-11 | |
|---|---|---|---|---|
| Priority: | Normal | Due date: | ||
| Assignee: | - | % Done: | 0% | |
| Category: | project described by recipe | Spent time: | - | |
| Target version: | - |
Description
long story short:
ANALYSIS-ERROR:
Error during analysis of #<PROJECT-SPEC lsp-csra-0.3:rsb-depth-sensors {1008C7B083}> Caused by:
> Error during analysis of /media/local_jenkins/tmp/project.euG0qW/branch-csra_0.3/ Caused by:
> > Could not resolve version "${Boost_USE_VERSION}" in find_package() expression.
ANALYSIS-ERROR:
Error during analysis of #<PROJECT-SPEC lsp-csra-0.3:rsb-depth-sensors {1008C7B083}> Caused by:
> Error during analysis of /media/local_jenkins/tmp/project.euG0qW/branch-csra_0.3/ Caused by:
> > Could not resolve version "${Boost_USE_VERSION}" in find_package() expression.
...
at several distributions on ayq: https://ayq.csra.cit-ec.de/view/DIST/job/Generate%20Distribution%200.3/lastBuild/console
History
#1
Updated by Jan Moringen over 8 years ago
- Category set to project described by recipe
Seems to be a bug in the project. The CMakeLists.txt file contains the following line:
find_package(Boost ${Boost_USE_VERSION} REQUIRED thread system filesystem)
but
Boost_USE_VERSION doesn't seem to be defined anywhere (as the generator remarks).
Compare to e.g. RSC which has the above line, but also:
SET(Boost_USE_VERSION 1.38 CACHE INTERNAL "Boost Version to use")
#2
Updated by Johannes Wienke over 8 years ago
This is a corner case, I would say. Despite having an undefined variable, the call, with the variable resolving to blank, is still valid. In theory this could be used to optionally let the user provide a version via the command line (ignoring the cache).
#3
Updated by Jan Moringen over 8 years ago
Johannes Wienke wrote:
This is a corner case, I would say. Despite having an undefined variable, the call, with the variable resolving to blank, is still valid. In theory this could be used to optionally let the user provide a version via the command line (ignoring the cache).
Sure. In this particular case however, I suspect that the find_package call has been copied from RSC without also copying the set call.
Also, the generator will still note the dependency, but without a version.
#4
Updated by Michael Goerlich over 8 years ago
both of you are right. I just got the impression, that jobs are not being updated (which was the result of a forgotten push actually). Maybe the word "warning" or "information" instead of "error" would be less misleading?
#5
Updated by Florian Lier over 7 years ago
- Status changed from New to Closed