| Age | Commit message (Collapse) | Author |
|---|
|
a tag we have the alternative of packaging/version content file (#5860)
|
|
script (#5806)
During the last major installer refactoring, we managed to break (again) the static64 installer.
We modified the code that defines user/group ownerships for netdata installation,
forcing the install to end up with a process running as netdata and files set as root
To fix this, we re-instate default NETDATA_USER/NETDATA_GROUP to root, as it should be.
Then we attempt group netadata creation. If that succeeds, we attempt user netdata creation.
If that succeeds, we attempt secondary groups addition.
Report errors on all otherwise situations from each step.
|
|
on the builds, not just make dist ^_^ (#5743)
|
|
refactoring wrap-up (#5736)
* remove dead quickfix
* unify global variable naming
* do not use double negation
* simplify setcap detection
* simplify logic of setting capabilities to apps.plugin
* always set group in chown
* simplify parameter parsing and banner notifications
* use built-in command for checking program availability
* lint functions.sh
* fix errors returned by shellcheck
* remove unused functions and use portable_service
* move user management functions closer together and remove "check" functions to reduce levels of indirection
* extract add_netdata_user_and_group and move it into installers for better code readability
* improve readability by not using global variable holding number of processors
* move netdata.conf file creation into one function
* revert migration to portable_service
* Less verbose setcap
* remove TODOs
* do not show output when not needed
* fix checking for group existence
* fix variable name
* netdata/packaging/installer: Dont spill out unnecessary output in stdout/stderr, it may confuse our users
The second commit was a follow up cleanup on nits.
* access: change codeowner
* Revert "access: change codeowner"
This reverts commit 6ee51ccae960584d5b1b7c243f5aabf046a99eff.
sorry, wrong repo
* netdata/packaging/installer: bug fix - do not use pidof as the name of the helper method
1) rename to safe_pidof, which is actually more accurate and describe what this method serves for (safe pid detection)
2) Renaming it also avoids the bug introduced by the usage of , that resulted in endless nested calls of the BASH method pidif
(Refer to command -v usage)
|
|
|
|
static build 64 breaks without it. Hasn't built properly since Dec.
|
|
|
|
|
|
* use 'git describe'
* no need to embed version string anymore
* fallback mechanism for package versioning
|
|
|
