3 ms·

> It will serve as a code style corpus to guide the process, and it's already a polished product, so you don't have to get into the nitty gritty of the last 20%

by GuiA 10y ago

> It will serve as a code style corpus to guide the process, and it's already a polished product, so you don't have to get into the nitty gritty of the last 20%. Yes, this is spot on. I work in an industrial R&D department where engineers are expected to be fairly independent; on a recent project, an engineer whose skillset we had misestimated during the interview had a really hard time starting the codebase from scratch, architecting the software, etc. It had to be saved close to the last minute by a senior engineer (to be fair, the manager should have kept a closer eye on the project). For that engineer's next project, we assigned her to something already established and architected, and she's been doing great. Her programming/architecture skills grew tremendously in a few months by being immersed in that codebase. (I was the senior engineer who had to be called in for the first project, and lead on the second project. The engineer in question was great in many other areas critical for the job, just her software carpentry skills that were lacking).

joatmon-snoo 10y ago
It's also a skill that schools don't do a very good job of. Right now I'm TA'ing a C++ class and the sheer stupidity of some of the questions that I get asked is ridiculous (I say stupidity, but it's really a mix of (1) some actual stupidity, (2) students not thinking for themselves, (3) students having no idea where to look and/or stuff was explained by hand-waving it away.) One thing we've done, for instance, is require that students submit their code via GitHub and make their code compile under both GCC and Clang on Travis. On the last assignment, the professor made a mistake when duplicating the Travis config file, so Travis kept trying to run a binary that wasn't there, because the new assignment had a different name for the compiled binary. The sheer number of emails and questions I got was absolutely ridiculous - and granted, I didn't blame the students for asking, because the entire toolchain side of things had sort of been hand-waved away from them, but the fix was literally replacing "./hw4-submission" with "./hw5-submission" in the Travis config. The point I'm getting at, I guess - because this isn't so much a thing about codebase architecture as knowing the tools you have to use - is that too much time is spent in the traditional CS education telling students to fill in code, or giving them frameworks and saying "make your thing work with this", instead of making them design stuff ground-up.
- voltagex_ 10y ago
  I don't have a lot of experience with Travis - was the error message straight forward? I wonder what a course teaching debugging from the error message backwards would look like? I was a semi-decent programmer before I started working but learning to debug took a lot more on-the-job WTFs to get the hang of.
  - joatmon-snoo 10y ago
    Sadly, no. It didn't help that there were a _lot_ of problems in the Travis/C++ ecosystem that complicated the matter. I actually have an outstanding PR into the Travis docs to update this which hasn't been merged yet for reasons unbeknownst to me https://github.com/travis-ci/docs-travis-ci-com/pull/746 https://github.com/travis-ci/docs-travis-ci-com/pull/746

by GuiA 10y ago

joatmon-snoo 10y ago

It's also a skill that schools don't do a very good job of. Right now I'm TA'ing a C++ class and the sheer stupidity of some of the questions that I get asked is ridiculous (I say stupidity, but it's really a mix of (1) some actual stupidity, (2) students not thinking for themselves, (3) students having no idea where to look and/or stuff was explained by hand-waving it away.) One thing we've done, for instance, is require that students submit their code via GitHub and make their code compile under both GCC and Clang on Travis. On the last assignment, the professor made a mistake when duplicating the Travis config file, so Travis kept trying to run a binary that wasn't there, because the new assignment had a different name for the compiled binary. The sheer number of emails and questions I got was absolutely ridiculous - and granted, I didn't blame the students for asking, because the entire toolchain side of things had sort of been hand-waved away from them, but the fix was literally replacing "./hw4-submission" with "./hw5-submission" in the Travis config. The point I'm getting at, I guess - because this isn't so much a thing about codebase architecture as knowing the tools you have to use - is that too much time is spent in the traditional CS education telling students to fill in code, or giving them frameworks and saying "make your thing work with this", instead of making them design stuff ground-up.

voltagex_ 10y ago
I don't have a lot of experience with Travis - was the error message straight forward? I wonder what a course teaching debugging from the error message backwards would look like? I was a semi-decent programmer before I started working but learning to debug took a lot more on-the-job WTFs to get the hang of.
- joatmon-snoo 10y ago
  Sadly, no. It didn't help that there were a _lot_ of problems in the Travis/C++ ecosystem that complicated the matter. I actually have an outstanding PR into the Travis docs to update this which hasn't been merged yet for reasons unbeknownst to me https://github.com/travis-ci/docs-travis-ci-com/pull/746 https://github.com/travis-ci/docs-travis-ci-com/pull/746