[elfabi] Add option to manually specify file read format

[amontanez]

Summary

Although llvm-elfabi will attempt to read input files without needing the format to be manually specified, doing so has the potential to introduce extraneous errors that can hinder debugging (since multiple readers may fail in attempts to read the file). This change allows the input file format to be manually specified to force elfabi to use a single reader. This makes it easier to test and debug errors specific to a given reader.

[jakehehrlich]

LGTM

[amontanez]

Small change to read-tbe-as-elf.test to make it easier to debug test failure.

[jhenderson]

Happy for this to be committed once my comments are addressed.

llvm/test/tools/llvm-elfabi/read-tbe-as-elf.test

15

Nit: I think the style is to have a space between the # and CHECK.

17

Maybe worth checking the filename part of this in this error message.

llvm/test/tools/llvm-elfabi/read-tbe-as-tbe.test

2

Just run FileCheck --input-file=%t, rather than using cat.

Not sure if this available here, but something worth considering adding if it isn't. A lot of tests put a file on stdout using '-' as the filename. Doing so would then allow you to run FileCheck on it without needing to call cat or use --input-file:
# RUN: llvm-elfabi -TBE %s -emit-tbe=- | FileCheck %s

llvm/tools/llvm-elfabi/llvm-elfabi.cpp

40

I take it there's no way of making the input argument match-up case insensitive? I know that options can be case-insensitive if the command line library is used appropriately.

[amontanez]

• Rebase on changes to parent patch.
• Fix test formatting.
• Input format flags are now lowercase only.

[amontanez]

llvm/tools/llvm-elfabi/llvm-elfabi.cpp

40

I looked into this for a while and couldn't find any way make these case insensitive. I assumed it would be possible too. I realized that the other command line options weren't case insensitive either, so I've set these to be lowercase only for consistency.

[jhenderson]

LGTM.

llvm/tools/llvm-elfabi/llvm-elfabi.cpp

40

@jakehehrlich, didn't you run into something like this in llvm-objcopy? Maybe you could shed some light on things.

[amontanez]

Closed in rL350545

llvm/test/tools/llvm-elfabi/read-tbe-as-elf.test

+# RUN: not llvm-elfabi -elf %s -emit-tbe=%t  2>&1 | FileCheck %s
+
+--- !tapi-tbe
+SoName: somelib.so
+TbeVersion: 1.0
+Arch: x86_64
+Symbols:
+  foo: { Type: Func }
+  bar: { Type: Object, Size: 42 }
+  baz: { Type: Object, Size: 8 }
+  not: { Type: Object, Undefined: true, Size: 128 }
+  nor: { Type: Func, Undefined: true }
+...
+
+#CHECK:         Error: Encountered multiple errors:
+#CHECK-NEXT:    0 (BinaryRead): The file was not recognized as a valid object file
+#CHECK-NEXT:    All file readers failed to read `{{.*}}`. (unsupported/malformed file?)

llvm/test/tools/llvm-elfabi/read-tbe-as-tbe.test

+# RUN: llvm-elfabi -TBE %s -emit-tbe=%t
+# RUN: cat %t | FileCheck %s
+
+--- !tapi-tbe
+TbeVersion: 1.0
+Arch: AArch64
+Symbols: {}
+...
+
+#CHECK:         --- !tapi-tbe
+#CHECK-NEXT:    TbeVersion: {{[1-9]\d*\.(0|([1-9]\d*))}}
+#CHECK-NEXT:    Arch: AArch64
+#CHECK-NEXT:    Symbols: {}
+#CHECK-NEXT:    ...

llvm/tools/llvm-elfabi/llvm-elfabi.cpp

 #include "llvm/Support/Errc.h"
 #include "llvm/Support/FileOutputBuffer.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include "llvm/Support/Path.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/TextAPI/ELF/TBEHandler.h"
 #include <string>
 
+namespace llvm {
+namespace elfabi {
+
+enum class FileFormat {
+  TBE,
+  ELF
+};
+
+} // end namespace elfabi
+} // end namespace llvm
+
 using namespace llvm;
 using namespace llvm::elfabi;
 
 // Command line flags:
+cl::opt<FileFormat> InputFileFormat(
+    cl::desc("Force input file format:"),
+    cl::values(clEnumValN(FileFormat::TBE,
+                          "TBE", "Read `input` as text-based ELF stub"),
+               clEnumValN(FileFormat::TBE,
+                          "tbe", "(alias for TBE)"),
+               clEnumValN(FileFormat::ELF,
+                          "ELF", "Read `input` as ELF binary"),
+               clEnumValN(FileFormat::ELF,
+                          "elf", "(alias for ELF)")));
 cl::opt<std::string> InputFilePath(cl::Positional, cl::desc("input"),
                                    cl::Required);
 cl::opt<std::string>
     EmitTBE("emit-tbe",
             cl::desc("Emit a text-based ELF stub (.tbe) from the input file"),
             cl::value_desc("path"));
 cl::opt<std::string> SetSoname(
     "so-name",
ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrError =
       MemoryBuffer::getFile(FilePath);
   if (!BufOrError) {
     return createStringError(BufOrError.getError(), "Could not open `%s`.",
                              FilePath.data());
   }
 
   std::unique_ptr<MemoryBuffer> FileReadBuffer = std::move(BufOrError.get());
   // First try to read as a binary (fails fast if not binary).
+  if (InputFileFormat.getNumOccurrences() == 0 ||
+      InputFileFormat == FileFormat::ELF) {
     Expected<std::unique_ptr<ELFStub>> StubFromELF =
         readELFFile(FileReadBuffer->getMemBufferRef());
     Error ELFReadError = StubFromELF.takeError();
     if (!ELFReadError) {
       return std::move(StubFromELF.get());
     }
     EC.addError(std::move(ELFReadError), "BinaryRead");
+  }
 
   // Fall back to reading as a tbe.
+  if (InputFileFormat.getNumOccurrences() == 0 ||
+      InputFileFormat == FileFormat::TBE) {
     Expected<std::unique_ptr<ELFStub>> StubFromTBE =
         readTBEFromBuffer(FileReadBuffer->getBuffer());
     Error TBEReadError = StubFromTBE.takeError();
     if (!TBEReadError) {
       return std::move(StubFromTBE.get());
     }
     EC.addError(std::move(TBEReadError), "YamlParse");
+  }
 
   // If both readers fail, build a new error that includes all information.
   EC.escalateToFatal();
   return EC.makeError();
 }
 
 int main(int argc, char *argv[]) {
   // Parse arguments.