md5-mismatch for every download

[swapneils]

Every time I try to load a system with this package, I'm running into an md5-mismatch error. When using the "retry evaluation" or "register local projects" restarts, the code runs fine.

Is there a mismatch between how quicklisp calculates md5 hashes and how we're calculating them in our own function?

[rudolfochrist]

Hmm,
we're using md5sum underneath. Can you run md5sum --version and paste it's output here. Can you paste a full backtrace as well? Currently I cannot reproduce this. Thank you.

[bo-tato]

What is your result of (ql-dist:enabled-dists)? I realized the md5 checking function is using ql-dist:find-release which is looping over enabled-dists and returning the first match. Maybe it is possible that quicklisp tries to install from one dist, but the md5 it gets from another. Like rudolfochrist says full backtrace of error and what quicklisp commands you're running would help a lot to reproduce and debug this.

[swapneils]

Installing md5sum in the command line resolved this issue, apparently mac systems don't have it by default. Aside from a README update I think we can close this ticket as non-actionable.

[bo-tato]

In #10 I changed it to use sb-md5 to compute md5sum if we are on sbcl. And if not on sbcl it will still shell out to md5sum, but should signal a clear error that Couldn't execute "md5sum": No such file or directory rather than just getting md5-mismatch later.

[rudolfochrist]

Installing md5sum in the command line resolved this issue, apparently mac systems don't have it by default.

Interesting. Perhaps we should adjust the install.sh script as well, because it uses sha256sum and I guess this also isn't on macos by default.

[bo-tato]

From what I can tell internet searching it seems on mac osx it should work both openssl: openssl dgst -sha256 FILENAME or shasum -a 256 FILENAME, both of which work also a basic install of debian and arch linux at least.