[PERF2BOLT] Do not issue parsing error on weird build ids

[rafauler]

Summary

In weird entries we were issueing a parse error. For example, in line 5 here:

6862acc063b0aa86595f52ff81628577df4296ff a.so
6862acc063b0aa86595f52ff81628577df4296ff a.so
6862acc063b0aa86595f52ff81628577df4296ff a.so
db758cb3c970044e78d5a4c99b011708a9995636 bin1
60326683eab31acfd03435d9ed4ff9a8 bin2
7d448e51851b4bdb33eac84f90e74628a14a5f00 b.so
742aa26e0211794356cc25f415c25230a26aa045 c.so

Error reading BOLT data input file: line 89, column 33: malformed field

Fix that.

[rafauler]

It is impossible to see in phabricator UI, but the line 5 I mention in the summary of this diff has several white spaces between a build ID and its binary name, and this build ID is shorter than all the other, so perf prints white spaces to align the binary name with the previous line.

[Amir]

In D134822#3821781, @rafauler wrote:

It is impossible to see in phabricator UI, but the line 5 I mention in the summary of this diff has several white spaces between a build ID and its binary name, and this build ID is shorter than all the other, so perf prints white spaces to align the binary name with the previous line.

You can try using markdown code block to preserve the original formatting.

bolt/lib/Profile/DataAggregator.cpp

DataAggregator::parseNameBuildIDPair() {
   }
 
   ErrorOr<StringRef> BuildIDStr = parseString(FieldSeparator, true);
   if (std::error_code EC = BuildIDStr.getError())
     return NoneType();
 
   // If one of the strings is missing, don't issue a parsing error, but still
   // do not return a value.
-  if (ParsingBuf[0] == '\n')
+  consumeAllRemainingFS();
+  if (checkAndConsumeNewLine())
     return NoneType();
 
   ErrorOr<StringRef> NameStr = parseString(FieldSeparator, true);
   if (std::error_code EC = NameStr.getError())
     return NoneType();
 
   consumeRestOfLine();
   return std::make_pair(NameStr.get(), BuildIDStr.get());