Conversation
* Add CMakeLists.txt to build geotrans library * Add CCI package configuration * Add functional test package using geotrans library provided sample code
* Build both libMSPdtcc and libMSPCoordinateConversionService to remain consistent with the upstream package. This is a change from the previous state in which both of these libraries were bundled into one. * Remove extraneous comments from test_package conanfile * Remove extraneous print statements from test_package example.cpp
* Add component definitions to package_info. This will allow cmake utilities like find_package to search for and find the package and its component libraries, which will enable linking against these individually for more granular control.
* No longer a shared library by default * Update homepage metadata * Update URL metadata
This comment has been minimized.
This comment has been minimized.
|
For the record: closes #1779 |
|
Thanks for your first submission. |
Co-authored-by: SpaceIm <30052553+SpaceIm@users.noreply.github.com> Co-authored-by: Anonymous Maarten <madebr@users.noreply.github.com>
* Fix CMake installation process * Refactor CMakeLists to be operated from the root source directory rather than source_subfolder. This removes the requirements to use the shutil package from within conanfile.py * Transition to usage of find_package for Threads library. This will allow thread library discovery to be performed in an OS agnostic way. * Test consumer package sets the required MPCCS_DATA environment variable to the data folder within the conan package folder rather than copying it to the consumers working directory. This is for CCI CI runner testing efficiency purposes, elimintating the need for a copy.
This comment has been minimized.
This comment has been minimized.
This will ideally make it easier for the user to understand the requirements of them to correctly configure the MPCCS_DATA environment variable prior to attempting to use the geotrans library.
This comment has been minimized.
This comment has been minimized.
* Delete fPIC option when shared library is being built * Remove superfluous shutil import
This comment has been minimized.
This comment has been minimized.
Co-authored-by: Rui Oliveira <ruimail24@gmail.com>
* Remove superfluous fPIC management from config_options function * Remove superfluous C library requirements to remove cpp flags
This comment has been minimized.
This comment has been minimized.
* Link MSPdtcc to MSPCoordinateConversionService to satisfy linkage requirements on macOS. This rectifies CI failures. * Remove cmake_find_package(_multi) assignments from conanfile.py at the advice of @madebr
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
* Move responsibility for setting MSPCCS_DATA environment variable from consumer to package * Remove usage output from build method * Remove explicit assignment of resdir value Co-authored-by: SpaceIm <30052553+SpaceIm@users.noreply.github.com>
This comment has been minimized.
This comment has been minimized.
* Remove version number of library from CMakeLists.txt * Remove version number of library from conanfile.py * Move definition of CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS from conanfile to CMakeLists.txt * Add commentary on requisite environment variables to be set to consumer conanfile.py * Re-add call to tools.environment_append to ensure that MPCCS_DATA environment variable is set when executing the example binary
This comment has been minimized.
This comment has been minimized.
|
I've written to the organization to see if it is possible to get a link with versioned sources... let's see if they answer in a couple of days. |
To follow this up, we received a response from NGA today. My understanding of the response is that because they're not interested in supporting legacy versions, they are not going to provide a versioned download link at this stage. They have indicated that they will consider the wider community need for a versioned download link and may revise this position in the future. My read of this is that they may be convinced to go to the effort of providing a download link if geotrans sees sufficient popularity on conan, if there are sufficient complaints about packages breaking when they release 3.9 (scheduled within the next 9 months), or perhaps just part of their 3.9 release process. In the meantime, it appears as though finding an alternate way to mirror the source code would be the best way forward. @jgsogo do you agree with my summary/have anything to add? |
|
Yes, same message. I'm proposing a PR here to store these sources (conan-io/cci-sources-backup#2). We cannot use regular Git because the file is too big, and (free) LFS storage/bandwith in GitHub is quite limited. We can configure LFS to use Artifactory, so we have plenty of room for these files. However, my main concern now is how to get the link from the repo. It requires some translation from to the actual URL: Random thoughts:
|
jgsogo
left a comment
There was a problem hiding this comment.
Please, use sources from the mirror in our Artifactory.
I realise that we're moving forward with a generic Artifactory repo in lieu of a LFS solution for the moment, but to provide some more discussion about why the Artifactory solution is preferred (at least from my perspective). There are practical considerations to using LFS that are messy. Namely, what is the desired workflow? Normally, this is:
There seem like two alternatives here:
I guess the Artifactory solution here is really the same as solution (2) above, as I would imagine only JFrog staff will have the ability to upload new sources (?). Regardless, I think the Artifactory solution simplifies things a bit as solution (1) becomes infeasible. The URL is simpler too. |
|
Let's KISS this. git is a handy tool, but we don't have to push it everywhere. |
* Update geotrans 3.8 source URL to use JFrog Artifactory. This resolves a build reproducibility issue caused by the fact that NGA don't host a versioned mirror of the source files. The JFrog mirror provided is now versioned and will be able to be downloaded even after the release of geotrans 3.9 Co-authored-by: Javier G. Sogo <jgsogo@gmail.com>
|
All green in build 15 (
|
jgsogo
left a comment
There was a problem hiding this comment.
🚀
Sorry for the delay and all the back and forth regarding the sources... but finally the package will be there. Thanks a lot for your patience!
KISS is exactly the reason, that's why we have chosen plain artifactory over artifactory + git lfs |

Specify library name and version: geotrans/3.8
I saw that there was an outstanding library request for it (#1779). I wanted to learn about conan packaging and this looked like a good candidate, since I also wanted to check the library out. I'm not the author of this library.
One of the hook tests is currently failing. I've raised #6511 for discussion of this issue and an appropriate way forward. I figured it deserved it's own issue as it raises questions about the assumptions of the "no imports" test more generally than just this package.closes #1779
conan-center hook activated.