Commit ba17f4b
committed
Check that
When retesting `gix-path` without `git` being in the `PATH`, this
adds a step to verify that the `EXEPATH` environment variable is
not set. If `EXEPATH` is set, the the `EXEPATH` optimization could
succeed even if the fallback strategy that is intended to be more
reliable would fail due to a regression.
This applies only to the `test-fast` jobs on Windows, where we
rerun `gix-path` tests (to make sure they work even when `git is
not in the path, when Git for Windows is installed in one of the
common locations). In contrast, fine if `EXEPATH` is set (or not)
in the `test-fixtures-windows` jobs.
Although it's possible to run the `gix-path` tests a third time
with `EXEPATH` set, it wouldn't really make sense to do so when
`git` is not found in `PATH`, since the scenarios where `EXEPATH`
is expected to have a usable value are in a Git Bash environment
where `git` will be in the `PATH` even if it is not otherwise.
(It might make sense to test a Git Bash environment explicitly, but
that would still be with `git` in `PATH`, and this does not add
such tests. It can be done later if needed to guard against a
regression, and it will also likely be done eventually if more Git
for Windows installations are tested: Git for Windows SDK, MinGit,
MinGit with Busybox, MinGit 32-bit, and/or Git for Windows running
inside a native Windows container. The absence of tests using a Git
Bash environment should not be confused with tests of interaction
with the Git for Windows provided `bash` shell on Windows. That we
do test: both directly, and indirectly but heavily in fixtures via
`gix-testtools`.)EXEPATH is unset on Windows CI1 parent 9000a84 commit ba17f4b
1 file changed
+3
-0
lines changed| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
246 | 246 | | |
247 | 247 | | |
248 | 248 | | |
| 249 | + | |
| 250 | + | |
| 251 | + | |
249 | 252 | | |
250 | 253 | | |
251 | 254 | | |
| |||
0 commit comments