You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am developing a CI flow where each affected dependent (calculated by lockfile build-order) is rebuilt by triggering it's respective pipeline job. I trigger the first layer, those jobs trigger the next layer, and so on, cascading all the way down to the end-product.
When all downstream jobs have finished, I update the original lockfile with the results and generate buildinfo. (I know buildinfo is not currently recommended, but even with the current issues it is still fit for my usage in package promotion.)
In order to generate build info I need to have a cache with all the packages, so I install from the updated lockfile:
but it fails because the install did not fetch the build requirements. I could not find any build policy which would make the build requirements without having to rebuild a package.
I think maybe a workaround would be to parse the lockfile to find the build requirements and use the experimental --lockfile-node-id LOCKFILE_NODE_ID options, but I'd appreciate a neater solution.
The text was updated successfully, but these errors were encountered:
I think this basically boils down to a feature request which is kind of related to this: #8121
Ideally, there would be a way to opt-into installing all the build_requires of a dependency graph or lockfile, even if not calling the build method.
There is already an extensive change to graph evaluation where the plan is to evaluate the whole graph on every operation (instead of the current logic, in which a bunch of things are skipped when build() is not being performed).
@memsharded , is the current plan for Conan 2.0 to always download/install build_requires, even if not building?
As a temporary workaround, I would try invoking conan info app1 --lockfile=complete_app1.lock --build which might retrieve at least the recipes (but not the binary).
Maybe also #8021 could also help with this need (this is work in progress as well).
I just ran into a similar issue, where such a parameter would be advantageous. What is the current recommended workaround to achieve downloading all build and normal requirements of a conan.py? (no previous commands were executed)
I am developing a CI flow where each affected dependent (calculated by lockfile build-order) is rebuilt by triggering it's respective pipeline job. I trigger the first layer, those jobs trigger the next layer, and so on, cascading all the way down to the end-product.
When all downstream jobs have finished, I update the original lockfile with the results and generate buildinfo. (I know buildinfo is not currently recommended, but even with the current issues it is still fit for my usage in package promotion.)
In order to generate build info I need to have a cache with all the packages, so I install from the updated lockfile:
conan install app1 --lockfile=complete_app1.lock
and generate the build info:
but it fails because the install did not fetch the build requirements. I could not find any build policy which would make the build requirements without having to rebuild a package.
I think maybe a workaround would be to parse the lockfile to find the build requirements and use the experimental
--lockfile-node-id LOCKFILE_NODE_ID
options, but I'd appreciate a neater solution.The text was updated successfully, but these errors were encountered: