| ??? 11/02/09 15:52 Read: times |
#170351 - Broken code in RTL is the worst Responding to: ???'s previous message |
Christoph Franck said:
Andy Neil said:
In most cases where code "works" without optimisation, but not with, it's because of a flaw in the code - not a bug in the compiler. That just makes the genuine compiler bugs even nastier. You'll spend hours/days/weeks combing through your own code until you eventually look at the generated assembly and find out the truth. Incorrect code generation is quite often easy to catch. The worst problems is unexpected bugs in the runtime library - especially if you don't have the source code, so you will have to reverse-engineer a call chain until you find the goof. |
| Topic | Author | Date |
| Odd calculation result... | 01/01/70 00:00 | |
| Hmm. Even odder. | 01/01/70 00:00 | |
| what is the definition... | 01/01/70 00:00 | |
| Compiler reuses registers and memory cells | 01/01/70 00:00 | |
| Ah. | 01/01/70 00:00 | |
| Disable compiler optimizations. | 01/01/70 00:00 | |
| Compiler bug? | 01/01/70 00:00 | |
| Maturity of compilers | 01/01/70 00:00 | |
| the primary suspect should always be the end user. | 01/01/70 00:00 | |
| "innovation" as alternative name for bug | 01/01/70 00:00 | |
| C51 also does that! | 01/01/70 00:00 | |
| C51 has smart linker | 01/01/70 00:00 | |
| I found a some genuine compiler bugs. | 01/01/70 00:00 | |
| Broken code in RTL is the worst | 01/01/70 00:00 | |
| Thanks! | 01/01/70 00:00 | |
| Defs | 01/01/70 00:00 | |
| poor choice of names | 01/01/70 00:00 | |
| Good data types are really critical | 01/01/70 00:00 | |
| I'm an idiot. | 01/01/70 00:00 | |
| informative! | 01/01/70 00:00 | |
| "Answer is wrong" | 01/01/70 00:00 | |
Which one would you prefer?? | 01/01/70 00:00 |