In my experience assertions are a development tool and should not be a part of the user's experience. With that in mind, do we expect the end user to ever invoke llc directly? The reason I ask is that I directly gave llc some bogus command line input and it asserted. Is it a bug that a graceful error message and exit from llc wasn't performed or is it considered "fair game" to generate an assertion because I didn't go through the larger tool chain? Jack -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20120809/42fa5427/attachment.html>
On Thu, Aug 9, 2012 at 1:00 PM, Carter, Jack <jcarter at mips.com> wrote:> In my experience assertions are a development tool and should not be a part > of the user's experience. > > With that in mind, do we expect the end user to ever invoke llc directly? > > The reason I ask is that I directly gave llc some bogus command line input > and it asserted. > > Is it a bug that a graceful error message and exit from llc wasn't performed > or is it considered "fair game" to generate an assertion because I didn't go > through the larger tool chain?Generally, it's a bug; if there's some construct we can't generate code for, we should be using report_fatal_error. Granted, bugs like that have much lower priority than bugs which can be triggered from realistic code. -Eli
Thanks ________________________________________ From: Eli Friedman [eli.friedman at gmail.com] Sent: Thursday, August 09, 2012 1:12 PM To: Carter, Jack Cc: List Subject: Re: [LLVMdev] Generic question about llc asserts On Thu, Aug 9, 2012 at 1:00 PM, Carter, Jack <jcarter at mips.com> wrote:> In my experience assertions are a development tool and should not be a part > of the user's experience. > > With that in mind, do we expect the end user to ever invoke llc directly? > > The reason I ask is that I directly gave llc some bogus command line input > and it asserted. > > Is it a bug that a graceful error message and exit from llc wasn't performed > or is it considered "fair game" to generate an assertion because I didn't go > through the larger tool chain?Generally, it's a bug; if there's some construct we can't generate code for, we should be using report_fatal_error. Granted, bugs like that have much lower priority than bugs which can be triggered from realistic code. -Eli
Reasonably Related Threads
- [LLVMdev] Generic question about llc asserts
- [LLVMdev] Using llvm-mc assembler in the llvm test-suite
- [LLVMdev] Using llvm-mc assembler in the llvm test-suite
- [LLVMdev] [MC] [llvm-mc] Getting target specific information to <target>ELFObjectWriter
- [LLVMdev] (MC) Register parsing for AsmParser (standalone assembler)