Debuggers Make You a Bad Detective(thinkingmuchbetter.com)
thinkingmuchbetter.com
Debuggers Make You a Bad Detective
https://www.thinkingmuchbetter.com/main/debugging-bad-detective/
9 comments
Depends really. If I've just joined a place where I've got no clue how the code works, the best thing I can do is use a debugger. It will make me understand how the code works, and will get me to the bugs quicker.
I agree though. If someone knows the code intimately and still doesn't have a rough clue about why a bug exists, and the first thing they do is look at the programs entry point and step into all the way until something pops out, it certainly will take you forever. And in this context, OP is certainly right.
I agree though. If someone knows the code intimately and still doesn't have a rough clue about why a bug exists, and the first thing they do is look at the programs entry point and step into all the way until something pops out, it certainly will take you forever. And in this context, OP is certainly right.
I tend to use ipython kernel for debugging in python, it is not a debugger but does the job for me to find what is wrong while working on a scientific project. What are your thoughts on it?
That's like a python notebook, right? Given the quick feedback loop, and that you're seeing just what you need to see, that seems like an efficient way to debug that sort of thing.
Yes, the jupyter notebook thing, rightly pointed out by you, it does give a quick feedback. One can also start writing code from the beginning in it and push a working cell into a .py file. This way one is very sure from the beginning that how the code will work. The downside of it, which I think, it feels to me that code is written in a very linear fashion.
Coming to debugger in python (pdb), in your blog you criticize using debugger to step line by line. But this need not be the case, one can jump to debugger prompt and set a break point from the prompt itself where one feels something is wrong and continue to that break point and repeat this process. So one need not to jump through each line. In such a case it works equivalently to carefully put print statements, i.e judiciously put print statements = judiciously put breakpoints, what do you think?
Coming to debugger in python (pdb), in your blog you criticize using debugger to step line by line. But this need not be the case, one can jump to debugger prompt and set a break point from the prompt itself where one feels something is wrong and continue to that break point and repeat this process. So one need not to jump through each line. In such a case it works equivalently to carefully put print statements, i.e judiciously put print statements = judiciously put breakpoints, what do you think?
I think a lot of people that criticize debuggers think professionals use them like a first year CS student, as a crutch because they don't understand the language they learning.
A professional uses them to trap events and then explore the programs state. And to inject faults and watch how the program explodes, or not. In that debuggers are vastly better than printf debugging.
A professional uses them to trap events and then explore the programs state. And to inject faults and watch how the program explodes, or not. In that debuggers are vastly better than printf debugging.
"a lot of people that criticize debuggers..." But what really making me think is that big giants like Torvalds, Rob Pike, Guido van Rossum. Rober C Martin, John Graham they don't like using debuggers (https://lemire.me/blog/2016/06/21/i-do-not-use-a-debugger).
That article is a great example: It defines "debugger" as "single-stepping through code", and then argues against that. It's fine to argue against "single-stepping through code" as a default technique, but artificially limiting the scope.
EDIT: one of the comments by the author makes it very clear: they don't consider lots of things you can do with e.g. gdb as "using a debugger": https://lemire.me/blog/2016/06/21/i-do-not-use-a-debugger/#c... </EDIT>
The submitted article here is similar, by opening with that ridiculous example and pretending that's what using a debugger means. Debuggers don't make you bad at finding problems. Not knowing when to use which tool makes you bad at it. (And knowing how to make new tools out of your existing ones is also quite helpful)
EDIT: one of the comments by the author makes it very clear: they don't consider lots of things you can do with e.g. gdb as "using a debugger": https://lemire.me/blog/2016/06/21/i-do-not-use-a-debugger/#c... </EDIT>
The submitted article here is similar, by opening with that ridiculous example and pretending that's what using a debugger means. Debuggers don't make you bad at finding problems. Not knowing when to use which tool makes you bad at it. (And knowing how to make new tools out of your existing ones is also quite helpful)
This is a good point, thank you! I am adding a quick note in the article to clarify that it's more about the code-stepping technique itself.
But gdb helped me understand why printf() was not working as intended, how those pesky register worked, remove memory leaks, and overall taught me a lot, fast. I had to rewrote malloc() using mmap at the start of my second year, i don't think any amount of printf() would've helped me as much as gdb and valgrind did. Also watchpoints basically do the same job as printf most of the time (but are more tedious to use).
I find pdb() usefull too (although it might be because i don't understand python as well as i should). I definitely use print() though, like said in the article, to narrow down where the break points should be placed. And i think than more than half the time i used pdb() it was on my test files, whether it was unit testing or feature testing (those are hard :/), so better test coverage would not help me much.
Also, debugging Common Lisp with (print) is usefull for beginners not used to the REPL, but once you understand how this whole slime/REPL thingy works, not using it to debug is bold.