[PATCH] Disable EHABI when -fno-exceptions option is chosen.

[Keith Walker]

A question to the more knowledgeable to try to clarify what is the expected
behaviour when compiling with EHABI exception handling enabled (i.e. the
current default for ARM) ....

 

Should the EHABI directives (.fnstart/.fnend/.push/.pad/.setfp etc) always
be generated (and thus resulting in the generation of .ARM.extab &
.ARM.exidx sections), or should these directives only be generated when
there are unwind tables to be described?

 

At the moment, even when compiling C code the .ARM.extab and .ARM.exidx
sections are being generated .... is this the desired behaviour?    This is
more what I would have expected if compiling C with -funwind-tables, but
maybe my expectations are wrong.

 

Keith

 

From: Renato Golin [mailto:renato.golin at linaro.org] 
Sent: 01 February 2014 23:35
To: Evgeniy Stepanov
Cc: reviews+D2666+public+b787a0c00c984fea at llvm-reviews.chandlerc.com; Alexey
Samsonov; Reid Kleckner; Keith Walker; Clang Commits; Joey Gouly
Subject: Re: [PATCH] Disable EHABI when -fno-exceptions option is chosen.

 

On 1 February 2014 19:17, Evgeniy Stepanov <eugenis at google.com> wrote:

We build sanitizer runtimes with exceptions disabled, because
otherwise we get a libstdc++ dependency, which is not a good thing.
Your last change effectively hardcodes the value of -arm-disable-ehabi
option - I can not override it on the command line, because now it is
always present in cc1 flags.

 

Ah, this is bad.

 

We can discuss more in the main list, maybe even jump directly to using
-munwind-tables instead. But for now let's play it safe. I'm at FOSDEM right
now, but feel free to revert it. If not, I'll do it on Monday. Sorry about
that. (some problems only show up when you break things ;)

 

cheers,

--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140203/35e19473/attachment.html>