[Git][ghc/ghc][wip/T18560] testsuite: Only run llvm ways if llc is available
Ben Gamari
gitlab at gitlab.haskell.org
Sun Aug 16 14:13:25 UTC 2020
Ben Gamari pushed to branch wip/T18560 at Glasgow Haskell Compiler / GHC
Commits:
27022c70 by Ben Gamari at 2020-08-16T10:13:19-04:00
testsuite: Only run llvm ways if llc is available
As noted in #18560, we previously would always run the LLVM ways since
`configure` would set `SettingsLlcCommand` to something non-null when
it otherwise couldn't find the `llc` executable. Now we rather probe for
the existence of the `llc` executable in the testsuite driver.
Fixes #18560.
- - - - -
4 changed files:
- hadrian/src/Settings/Builders/RunTest.hs
- testsuite/config/ghc
- testsuite/driver/testglobals.py
- testsuite/mk/test.mk
Changes:
=====================================
hadrian/src/Settings/Builders/RunTest.hs
=====================================
@@ -70,7 +70,6 @@ runTestBuilderArgs = builder RunTest ? do
withSMP <- getBooleanSetting TestGhcWithSMP
debugged <- getBooleanSetting TestGhcDebugged
keepFiles <- expr (testKeepFiles <$> userSetting defaultTestArgs)
- withLlvm <- expr (not . null <$> settingsFileSetting SettingsFileSetting_LlcCommand)
accept <- expr (testAccept <$> userSetting defaultTestArgs)
(acceptPlatform, acceptOS) <- expr . liftIO $
@@ -125,8 +124,6 @@ runTestBuilderArgs = builder RunTest ? do
, arg "-e", arg $ asBool "config.have_profiling=" (hasLibWay profiling)
, arg "-e", arg $ asBool "config.have_fast_bignum=" (bignumBackend /= "native" && not bignumCheck)
, arg "-e", arg $ asBool "ghc_with_smp=" withSMP
- , arg "-e", arg $ asBool "ghc_with_llvm=" withLlvm
-
, arg "-e", arg $ "config.ghc_dynamic_by_default=" ++ show hasDynamicByDefault
, arg "-e", arg $ "config.ghc_dynamic=" ++ show hasDynamic
=====================================
testsuite/config/ghc
=====================================
@@ -64,10 +64,6 @@ else:
if (config.have_profiling and ghc_with_threaded_rts):
config.run_ways.append('profthreaded')
-if (ghc_with_llvm and not config.unregisterised):
- config.compile_ways.append('optllvm')
- config.run_ways.append('optllvm')
-
# WinIO I/O manager for Windows
if windows:
winio_ways = ['winio', 'winio_threaded']
@@ -195,6 +191,17 @@ def get_compiler_info():
config.have_ncg = compilerInfoDict.get("Have native code generator", "NO") == "YES"
+ # Detect whether an LLVM toolhain is available
+ llc_path = compilerInfoDict.get("LLVM llc command")
+ config.have_llvm = shutil.which(llc_path) is not None
+ if config.unregisterised:
+ print("Unregisterised build; skipping LLVM ways...")
+ elif config.have_llvm:
+ config.compile_ways.append('optllvm')
+ config.run_ways.append('optllvm')
+ else:
+ print("Failed to find `llc` command; skipping LLVM ways...")
+
# Whether GHC itself was built using the LLVM backend. We need to know this
# since some tests in ext-interp fail when stage2 ghc is built using
# LLVM. See #16087.
=====================================
testsuite/driver/testglobals.py
=====================================
@@ -115,6 +115,9 @@ class TestConfig:
self.way_flags = {} # type: Dict[WayName, List[str]]
self.way_rts_flags = {} # type: Dict[WayName, List[str]]
+ # Do we have a functional LLVM toolchain?
+ self.have_llvm = False
+
# Do we have vanilla libraries?
self.have_vanilla = False
=====================================
testsuite/mk/test.mk
=====================================
@@ -194,18 +194,6 @@ else
RUNTEST_OPTS += -e ghc_with_smp=False
endif
-# Does the LLVM backend work?
-ifeq "$(LLC)" ""
-RUNTEST_OPTS += -e ghc_with_llvm=False
-else ifeq "$(TargetARCH_CPP)" "powerpc"
-RUNTEST_OPTS += -e ghc_with_llvm=False
-else ifneq "$(LLC)" "llc"
-# If we have a real detected value for LLVM, then it really ought to work
-RUNTEST_OPTS += -e ghc_with_llvm=True
-else
-RUNTEST_OPTS += -e ghc_with_llvm=False
-endif
-
ifeq "$(WINDOWS)" "YES"
RUNTEST_OPTS += -e windows=True
else
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/27022c70a30cfff7a3b17817b1e33ab302cd5ecf
--
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/-/commit/27022c70a30cfff7a3b17817b1e33ab302cd5ecf
You're receiving this email because of your account on gitlab.haskell.org.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-commits/attachments/20200816/29033bbe/attachment-0001.html>
More information about the ghc-commits
mailing list