[llvm][support] Fix ScopedPrinterTest on AIX

[paulkirth]

Summary

The test strings we used for infinity and NAN were not correct on AIX.
This patch creates those dynamically instead of hard-coded.

[abhina.sreeskantharajan]

LGTM, thanks for the quick fix!

[jhenderson]

llvm/unittests/Support/ScopedPrinterTest.cpp

598

Dumb question maybe, but why the "5.1"?

[paulkirth]

llvm/unittests/Support/ScopedPrinterTest.cpp

598

"%5.1" is the format string we use in printNumber(). IIRC that was what was taken from existing code in ElfDumper.

[abhina.sreeskantharajan]

Hi @paulkirth, this testcase is still failing the same way on the AIX bot even after this change https://lab.llvm.org/buildbot/#/builders/214/builds/6557. Can this change be reverted until we figure out the proper fix?

[paulkirth]

Ugh, I see this issue. I missed that the "INF/inf" discrepancy also happens for JSON output, but only changed the non-JSON test string. I should have a forward fix very soon. If that causes issues, we'll either need to revert the original patch, the 2 fixes and the HasHistogram patch, or disable the test for float/double,until we sort it out.

If you really need the bot to progress, I'm fine w/ you reverting things, though.

[paulkirth]

@abhina.sreeskantharajan is there a way to test AIX things locally? i.e. as in run on a bot "somewhere"? or a simple QEMU image we can use, etc?

[paulkirth]

also, forward fix in https://reviews.llvm.org/D146739

[abhina.sreeskantharajan]

In D146542#4216973, @paulkirth wrote:

Ugh, I see this issue. I missed that the "INF/inf" discrepancy also happens for JSON output, but only changed the non-JSON test string. I should have a forward fix very soon. If that causes issues, we'll either need to revert the original patch, the 2 fixes and the HasHistogram patch, or disable the test for float/double,until we sort it out.

If you really need the bot to progress, I'm fine w/ you reverting things, though.

Thanks, I see you already put up a fix so we can see if that fixes it

In D146542#4217072, @paulkirth wrote:

@abhina.sreeskantharajan is there a way to test AIX things locally? i.e. as in run on a bot "somewhere"? or a simple QEMU image we can use, etc?

I believe you can get access to AIX systems in GCC Compile Farm system http://gcc.gnu.org/wiki/CompileFarm

llvm/unittests/Support/ScopedPrinterTest.cpp

 //===- llvm/unittest/Support/ScopedPrinterTest.cpp - ScopedPrinter tests --===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Support/ScopedPrinter.h"
 #include "llvm/ADT/APSInt.h"
+#include "llvm/Support/Format.h"
 #include "gtest/gtest.h"
 #include <cmath>
 #include <vector>
 
 using namespace llvm;
 
 TEST(JSONScopedPrinterTest, PrettyPrintCtor) {
   auto PrintFunc = [](ScopedPrinter &W) {
TEST_F(ScopedPrinterTest, PrintNumber) {
   char Buf[512];
 
   format("%5.1f", std::numeric_limits<float>::max()).snprint(Buf, sizeof(Buf));
   std::string MaxFloatStr(Buf);
 
   format("%5.1f", std::numeric_limits<double>::max()).snprint(Buf, sizeof(Buf));
   std::string MaxDoubleStr(Buf);
 
+  format("%5.1f", std::numeric_limits<double>::infinity())
+      .snprint(Buf, sizeof(Buf));
+  std::string InfFloatStr(Buf);
+
+  std::to_string(std::numeric_limits<float>::infinity());
+  std::string InfDoubleStr(Buf);
+
+  format("%5.1f", std::nanf("1")).snprint(Buf, sizeof(Buf));
+  std::string NaNFloatStr(Buf);
+
+  format("%5.1f", std::nan("1")).snprint(Buf, sizeof(Buf));
+  std::string NaNDoubleStr(Buf);
+
   std::string ExpectedOut = Twine(
                                 R"(uint64_t-max: 18446744073709551615
 uint64_t-min: 0
 uint32_t-max: 4294967295
 uint32_t-min: 0
 uint16_t-max: 65535
 uint16_t-min: 0
 uint8_t-max: 255
 uint8_t-min: 0
 int64_t-max: 9223372036854775807
 int64_t-min: -9223372036854775808
 int32_t-max: 2147483647
 int32_t-min: -2147483648
 int16_t-max: 32767
 int16_t-min: -32768
 int8_t-max: 127
 int8_t-min: -128
 apsint: 9999999999999999999999
 label: value (0)
 float-max: )" + MaxFloatStr + R"(
 float-min:   0.0
-float-inf:   inf
-float-nan:   nan
+float-inf: )" + InfFloatStr + R"(
+float-nan: )" + NaNFloatStr + R"(
 float-42.0:  42.0
 float-42.5625:  42.6
 double-max: )" + MaxDoubleStr +
                                 R"(
 double-min:   0.0
-double-inf:   inf
-double-nan:   nan
+double-inf: )" + InfDoubleStr + R"(
+double-nan: )" + NaNDoubleStr + R"(
 double-42.0:  42.0
 double-42.5625:  42.6
 )")
                                 .str();
 
   const char *JSONExpectedOut = R"({
   "uint64_t-max": 18446744073709551615,
   "uint64_t-min": 0,