This is the mail archive of the
frysk@sources.redhat.com
mailing list for the frysk project.
Re: [patch request] don't run ecj on auto generated files
- From: Andrew Cagney <cagney at redhat dot com>
- To: Stan Cox <scox at redhat dot com>
- Cc: frysk at sources dot redhat dot com
- Date: Fri, 28 Jul 2006 09:50:48 -0400
- Subject: Re: [patch request] don't run ecj on auto generated files
- References: <44C94D19.3090803@redhat.com>
Stan,
I'm not personally warming to the idea for a number of reasons: one of
the key reasons for putting the code through ECJ is that that compiler
detects bad source that GCJ will accept and turn into bad code - even
generated files are exposed to this risk; very shortly GCJ will switch
to ECJ as it's front end and we'll be right back where we started; and
too often a warning, even in generated code, indicates that there is a
problem, even if it is with the generator (the other generated files
don't have problems right?).
Since antlr is the real concern, can something more local be done? For
instance, is there a newer antlr and is it any better? Or can the
antlr warnings be handled another way? Howe aware is upstream of the
issues?
Andrew
antlr generates files which generate warnings from ecj, for example
unreferenced variables. The current strategy is to use sed scripts to
postprocess the sources. However, this is error prone and requires
constant tweaking. Are there any objections to only running ecj on
top_srcdir *.java files and not running ecj on top_builddir *.java
files? I think this change would mostly effect antlr (Cpp*), but here
is the entire list of autogenerated files:
AT.java ELF.java SHF.java
Build.java EM.java SHN.java
Config.java ET.java SHT.java
CppLexer.java EV.java Sig.java
CppParser.java JUnitTests.java STB.java
CppParserTokenTypes.jav NT.java STT.java
CppTreeParser.java PF.java STV.java
CppTreeParserTokenTypes PowerPC.java SyscallNum.java
DW.java PT.java TestRunner.java
EI.java