appveyor: Downgrade MinGW to 6.2.0

It looks like the 6.3.0 MinGW comes with a gdb which has issues (#40184) that an
attempted workaround (#40777) does not actually fix (#40835). The original
motivation for upgradin MinGW was to fix build flakiness (#40546) due to newer
builds not exhibiting the same bug, so let's hope that 6.2.0 isn't too far back
in time and still contains the fix we need.

Closes #40835

appveyor.yml

@@ -46,13 +46,13 @@ environment:
     RUST_CONFIGURE_ARGS: --build=i686-pc-windows-gnu --enable-ninja
     SCRIPT: python x.py test
     MINGW_URL: https://s3.amazonaws.com/rust-lang-ci/rust-ci-mirror
-    MINGW_ARCHIVE: i686-6.3.0-release-win32-dwarf-rt_v5-rev1.7z
+    MINGW_ARCHIVE: i686-6.2.0-release-win32-dwarf-rt_v5-rev1.7z
     MINGW_DIR: mingw32
   - MSYS_BITS: 64
     SCRIPT: python x.py test
     RUST_CONFIGURE_ARGS: --build=x86_64-pc-windows-gnu --enable-ninja
     MINGW_URL: https://s3.amazonaws.com/rust-lang-ci/rust-ci-mirror
-    MINGW_ARCHIVE: x86_64-6.3.0-release-win32-seh-rt_v5-rev1.7z
+    MINGW_ARCHIVE: x86_64-6.2.0-release-win32-seh-rt_v5-rev1.7z
     MINGW_DIR: mingw64
 
   # 32/64 bit MSVC and GNU deployment
@@ -71,14 +71,14 @@ environment:
     RUST_CONFIGURE_ARGS: --build=i686-pc-windows-gnu --enable-extended --enable-ninja
     SCRIPT: python x.py dist
     MINGW_URL: https://s3.amazonaws.com/rust-lang-ci/rust-ci-mirror
-    MINGW_ARCHIVE: i686-6.3.0-release-win32-dwarf-rt_v5-rev1.7z
+    MINGW_ARCHIVE: i686-6.2.0-release-win32-dwarf-rt_v5-rev1.7z
     MINGW_DIR: mingw32
     DEPLOY: 1
   - MSYS_BITS: 64
     SCRIPT: python x.py dist
     RUST_CONFIGURE_ARGS: --build=x86_64-pc-windows-gnu --enable-extended --enable-ninja
     MINGW_URL: https://s3.amazonaws.com/rust-lang-ci/rust-ci-mirror
-    MINGW_ARCHIVE: x86_64-6.3.0-release-win32-seh-rt_v5-rev1.7z
+    MINGW_ARCHIVE: x86_64-6.2.0-release-win32-seh-rt_v5-rev1.7z
     MINGW_DIR: mingw64
     DEPLOY: 1
 

src/tools/compiletest/src/procsrv.rs

@@ -58,24 +58,8 @@ pub fn run(lib_path: &str,
     let mut cmd = Command::new(prog);
     cmd.args(args)
         .stdout(Stdio::piped())
-        .stderr(Stdio::piped());
-
-    // Why oh why do we sometimes make a pipe and sometimes inherit the stdin
-    // stream, well that's an excellent question! In theory it should suffice to
-    // always create a pipe here and be done with it. Unfortunately though
-    // there's apparently something odd with the gdb that comes with gcc 6.3.0
-    // on MinGW. Tracked at rust-lang/rust#40184 when stdin is piped here
-    // (unconditionally) then all gdb tests will fail on MinGW when using gcc
-    // 6.3.0. WHen using an inherited stdin though they happen to all work!
-    //
-    // As to why this fixes the issue, well, I have no idea. If you can remove
-    // this branch and unconditionally use `piped` and it gets past @bors please
-    // feel free to send a PR!
-    if input.is_some() || !cfg!(windows) {
-        cmd.stdin(Stdio::piped());
-    } else {
-        cmd.stdin(Stdio::inherit());
-    }
+        .stderr(Stdio::piped())
+        .stdin(Stdio::piped());
 
     add_target_env(&mut cmd, lib_path, aux_path);
     for (key, val) in env {