The pool builder tries to be minimal so it's fine for present/locked
packages not be assigned a solver/pool id. Adding a test to verify
correct creation of uninstall jobs
These special commands no longer (ab)use the partial update mechanism
but rather create a special install request for all current lock file
contents and later override any modified code references to the
originals. This leads to up to date remote metadata but no other
changes.
Commit: 149250ab92
ProcessExecutor::escape handled a false value inconsistently across
platforms, returning an emtpy string on Windows, otherwise `''`. This
is fixed to return `""` on Windows.
The GitDownloaderTest code has been appropriately updated.
Only when a install directory was not specified, was the CWD prepended to `$directory`. This change provides consistency in paths displayed to the user.
Provides feedback output before a potentially long wait on getBestCandidate() call on slow network connections where unresponsiveness/hang may be assumed.
Avoid waiting until after `getBestCandidate()` has finished, as it can add notably delay on slow connections due to downloading megabytes of data. Only to fail if the install location is invalid.
This only removes the credentials if they are managed by composer auth.json or equivalent, if the credentials were present in the package URL to begin with they might remain
Refs #8293Fixes#3644Closes#3608
Particularly the test
tests/Composer/Test/Fixtures/installer/partial-update-downgrades-non-whitelisted-unstable.test
is interesting because it verifies that an older version will be
installed on update if the new one is only present in the installed repo
or vendor dir. This was the cause of a lot of weird edge cases and
unreliable update behavior in Composer v1
As a result some lock file packages are no longer in the pool, so the
former installed map, now present map cannot use package ids anymore
Need to revisit some more code later to simplify this, todo notes left
Ensures packages get loaded from locked repo correctly. We may not want
to support this particular use-case at all, but for now it fixes the
existing test, so we may want to revisit this later.
Nils Adermann