Rollup merge of rust-lang#81734 - richkadel:fixfordash, r=pnkfelix

Ensures `make` tests run under /bin/dash (if available), like CI, and fixes a Makefile

Note: This cherrypicks rust-lang#81688 (`@pnkfelix)`

Updates `tools.mk` to explicitly require `SHELL := /bin/dash`, since CI uses `dash` but other environments (including developer local machines) may default to `bash`.

Replaces bash-specific shell command in one Makefile with a dash-compatible alternative, and re-enables the affected Makefile test.

Removes apparently redundant definition of `UNAME`.

Also see: [zulip discussion thread](https://rust-lang.zulipchat.com/#narrow/stream/131828-t-compiler/topic/how.20to.20run.2Fbless.20src.2Ftest.2Frun-make-fulldeps.2Fcoverage.20.3F)

r? `@pnkfelix`

FYI: `@wesleywiser` `@tmandry`

src/test/run-make-fulldeps/coverage-reports/Makefile

@@ -1,4 +1,3 @@
-# ignore-test Broken; accidentally silently ignored on Linux CI; FIXME(#81688)
 # needs-profiler-support
 # ignore-windows-gnu
 # min-llvm-version: 11.0
@@ -128,7 +127,7 @@ endif
 			$$( \
 				for file in $(TMPDIR)/rustdoc-$@/*/rust_out; \
 				do \
-				[[ -x $$file ]] && printf "%s %s " -object $$file; \
+				[ -x "$$file" ] && printf "%s %s " -object $$file; \
 				done \
 			) \
 		2> "$(TMPDIR)"/show_coverage_stderr.$@.txt \

src/test/run-make-fulldeps/coverage-reports/expected_show_coverage.partial_eq.txt

@@ -2,11 +2,11 @@
     2|       |// structure of this test.
     3|       |
     4|      2|#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord)]
-                       ^0            ^0      ^0 ^0  ^1       ^0 ^0^0
+                       ^0            ^0      ^0 ^0  ^1       ^1 ^0^0
     5|       |pub struct Version {
     6|       |    major: usize,
-    7|      1|    minor: usize, // Count: 1 - `PartialOrd` compared `minor` values in 3.2.1 vs. 3.3.0
-    8|      0|    patch: usize, // Count: 0 - `PartialOrd` was determined by `minor` (2 < 3)
+    7|       |    minor: usize,
+    8|       |    patch: usize,
     9|       |}
    10|       |
    11|       |impl Version {
@@ -45,19 +45,4 @@
    44|       |`function_source_hash` without a code region, if necessary.
    45|       |
    46|       |*/
-   47|       |
-   48|       |// FIXME(#79626): The derived traits get coverage, which is great, but some of the traits appear
-   49|       |// to get two coverage execution counts at different positions:
-   50|       |//
-   51|       |// ```text
-   52|       |//    4|      2|#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord)]
-   53|       |//                       ^0            ^0      ^0 ^0  ^1       ^0 ^0^0
-   54|       |// ```text
-   55|       |//
-   56|       |// `PartialEq`, `PartialOrd`, and `Ord` (and possibly `Eq`, if the trait name was longer than 2
-   57|       |// characters) have counts at their first and last characters.
-   58|       |//
-   59|       |// Why is this? Why does `PartialOrd` have two values (1 and 0)? This must mean we are checking
-   60|       |// distinct coverages, so maybe we don't want to eliminate one of them. Should we merge them?
-   61|       |// If merged, do we lose some information?
 

src/test/run-make-fulldeps/coverage/coverage_tools.mk

@@ -12,5 +12,3 @@
 # Enabling `-C link-dead-code` is not necessary when compiling with `-Z instrument-coverage`,
 # due to improvements in the coverage map generation, to add unreachable functions known to Rust.
 # Therefore, `-C link-dead-code` is no longer automatically enabled.
-
-UNAME = $(shell uname)

src/test/run-make-fulldeps/coverage/partial_eq.rs

@@ -4,8 +4,8 @@
 #[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord)]
 pub struct Version {
     major: usize,
-    minor: usize, // Count: 1 - `PartialOrd` compared `minor` values in 3.2.1 vs. 3.3.0
-    patch: usize, // Count: 0 - `PartialOrd` was determined by `minor` (2 < 3)
+    minor: usize,
+    patch: usize,
 }
 
 impl Version {
@@ -44,18 +44,3 @@ one expression, which is allowed, but the `function_source_hash` was only passed
 `function_source_hash` without a code region, if necessary.
 
 */
-
-// FIXME(#79626): The derived traits get coverage, which is great, but some of the traits appear
-// to get two coverage execution counts at different positions:
-//
-// ```text
-//    4|      2|#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord)]
-//                       ^0            ^0      ^0 ^0  ^1       ^0 ^0^0
-// ```text
-//
-// `PartialEq`, `PartialOrd`, and `Ord` (and possibly `Eq`, if the trait name was longer than 2
-// characters) have counts at their first and last characters.
-//
-// Why is this? Why does `PartialOrd` have two values (1 and 0)? This must mean we are checking
-// distinct coverages, so maybe we don't want to eliminate one of them. Should we merge them?
-// If merged, do we lose some information?

src/test/run-make-fulldeps/tools.mk

@@ -21,6 +21,19 @@ CGREP := "$(S)/src/etc/cat-and-grep.sh"
 # diff with common flags for multi-platform diffs against text output
 DIFF := diff -u --strip-trailing-cr
 
+# Some of the Rust CI platforms use `/bin/dash` to run `shell` script in
+# Makefiles. Other platforms, including many developer platforms, default to
+# `/bin/bash`. (In many cases, `make` is actually using `/bin/sh`, but `sh`
+# is configured to execute one or the other shell binary). `dash` features
+# support only a small subset of `bash` features, so `dash` can be thought of as
+# the lowest common denominator, and tests should be validated against `dash`
+# whenever possible. Most developer platforms include `/bin/dash`, but to ensure
+# tests still work when `/bin/dash`, if not available, this `SHELL` override is
+# conditional:
+ifneq (,$(wildcard /bin/dash))
+SHELL := /bin/dash
+endif
+
 # This is the name of the binary we will generate and run; use this
 # e.g. for `$(CC) -o $(RUN_BINFILE)`.
 RUN_BINFILE = $(TMPDIR)/$(1)