r344375 - [Driver] check for exit code from SIGPIPE

[Nick Desaulniers via cfe-commits]

Thanks for the quick revert to keep the build green; sorry for breaking it.
Maybe we can discuss a fix in:
https://reviews.llvm.org/D53001
On Fri, Oct 12, 2018 at 11:11 AM Reid Kleckner <rnk at google.com> wrote:
>
> This patch makes a lot of posix assumptions, and that header file doesn't exist on Windows. I reverted this to fix the build, but you might want to reconsider how this is designed.
>
> On Fri, Oct 12, 2018 at 10:24 AM Nick Desaulniers via cfe-commits <cfe-commits at lists.llvm.org> wrote:
>>
>> Author: nickdesaulniers
>> Date: Fri Oct 12 10:22:46 2018
>> New Revision: 344375
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=344375&view=rev
>> Log:
>> [Driver] check for exit code from SIGPIPE
>>
>> Summary:
>> D53000 adds a special exit code for SIGPIPE (writing to a closed
>> reader), and rather than print a fatal warning, skips printing the
>> error.  This can be seen commonly from piping into head, tee, or
>> split.
>>
>> Fixes PR25349, rdar://problem/14285346, b/77310947.
>>
>> Reviewers: jfb
>>
>> Reviewed By: jfb
>>
>> Subscribers: cfe-commits, thakis, srhines
>>
>> Differential Revision: https://reviews.llvm.org/D53001
>>
>> Modified:
>>     cfe/trunk/lib/Driver/Driver.cpp
>>
>> Modified: cfe/trunk/lib/Driver/Driver.cpp
>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Driver/Driver.cpp?rev=344375&r1=344374&r2=344375&view=diff
>> ==============================================================================
>> --- cfe/trunk/lib/Driver/Driver.cpp (original)
>> +++ cfe/trunk/lib/Driver/Driver.cpp Fri Oct 12 10:22:46 2018
>> @@ -78,6 +78,7 @@
>>  #include "llvm/Support/raw_ostream.h"
>>  #include <map>
>>  #include <memory>
>> +#include <sysexits.h>
>>  #include <utility>
>>  #if LLVM_ON_UNIX
>>  #include <unistd.h> // getpid
>> @@ -1388,8 +1389,9 @@ int Driver::ExecuteCompilation(
>>
>>    // Otherwise, remove result files and print extra information about abnormal
>>    // failures.
>> +  int Res = 0;
>>    for (const auto &CmdPair : FailingCommands) {
>> -    int Res = CmdPair.first;
>> +    int CommandRes = CmdPair.first;
>>      const Command *FailingCommand = CmdPair.second;
>>
>>      // Remove result files if we're not saving temps.
>> @@ -1398,10 +1400,17 @@ int Driver::ExecuteCompilation(
>>        C.CleanupFileMap(C.getResultFiles(), JA, true);
>>
>>        // Failure result files are valid unless we crashed.
>> -      if (Res < 0)
>> +      if (CommandRes < 0)
>>          C.CleanupFileMap(C.getFailureResultFiles(), JA, true);
>>      }
>>
>> +    // llvm/lib/Support/Unix/Signals.inc will exit with a special return code
>> +    // for SIGPIPE. Do not print diagnostics for this case.
>> +    if (CommandRes == EX_IOERR) {
>> +      Res = CommandRes;
>> +      continue;
>> +    }
>> +
>>      // Print extra information about abnormal failures, if possible.
>>      //
>>      // This is ad-hoc, but we don't want to be excessively noisy. If the result
>> @@ -1411,17 +1420,17 @@ int Driver::ExecuteCompilation(
>>      // diagnostics, so always print the diagnostic there.
>>      const Tool &FailingTool = FailingCommand->getCreator();
>>
>> -    if (!FailingCommand->getCreator().hasGoodDiagnostics() || Res != 1) {
>> +    if (!FailingCommand->getCreator().hasGoodDiagnostics() || CommandRes != 1) {
>>        // FIXME: See FIXME above regarding result code interpretation.
>> -      if (Res < 0)
>> +      if (CommandRes < 0)
>>          Diag(clang::diag::err_drv_command_signalled)
>>              << FailingTool.getShortName();
>>        else
>> -        Diag(clang::diag::err_drv_command_failed) << FailingTool.getShortName()
>> -                                                  << Res;
>> +        Diag(clang::diag::err_drv_command_failed)
>> +            << FailingTool.getShortName() << CommandRes;
>>      }
>>    }
>> -  return 0;
>> +  return Res;
>>  }
>>
>>  void Driver::PrintHelp(bool ShowHidden) const {
>>
>>
>> _______________________________________________
>> cfe-commits mailing list
>> cfe-commits at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits



-- 
Thanks,
~Nick Desaulniers