Custom Query (1630 matches)

If you try to install md5 twice in a row, it fails with an error:

$ ~/chickens/5.3.0rc4/bin/chicken-install -test md5
building md5
/home/sjamaan/.cache/chicken-install/md5/md5.h: No such file or directory

Error: shell command terminated with nonzero exit code
256
"sh /home/sjamaan/.cache/chicken-install/md5/md5.build.sh"

This also happens with 5.2.0, and also if you just build but not install (i.e., chicken-install -n md5).

Any idea why this happens? Is this a bug in chicken-install or in the egg?

The md5 egg's test suite will fail if you call chicken-install -test md5 when there is no csi binary in $PATH. I tried this just now while testing 5.3.0rc4 after installing it into a temporary directory.

This means if the CHICKEN binaries *are* in $PATH but uses PROGRAM_PREFIX or PROGRAM_SUFFIX it won't work either.

Worse, if the csi in $PATH is a different one from the one corresponding to the chicken-install being called, the results may be misleading - either failure if the one installed there is buggy, or passing tests if the one installed there works even if the egg that was just installed is buggy.

I notice that there's a fallback to use $CHICKEN_CSI and $CHICKEN_CSC, but it appears that chicken-install only passes those options to the build script, not the test script.

run.scm is already loaded from the correct csi, so for the interpreter I'd suggest just loading/including the test suite. But perhaps it's a good idea if chicken-install would export those same environment variables when invoking the test suite... And/or have some way to figure out how to call the chicken binaries by exposing them from a module? (but that would be backwards incompatible, so the test suite would need to deal with older CHICKEN versions gracefully anyway).

If you perform chicken-install srfi-41 and inspect the generated streams.derived.link file, it contains srfi-9 as a dependency, which is incorrect because srfi-9 is a primitive module which is part of libchicken, so no static linking is needed.

This breaks anything that relies on srfi-41 and tries to link statically, for example ​https://github.com/markjfisher/aoc-chicken/blob/master/build/static-build.sh

This seems to be a regression from 5.2.0 (and therefore a definite blocker for 5.3.0), because there the generated link file contains only (streams.primitive type-checks), which is correct.

I haven't done any serious investigation yet, but likely candidates for this regression are 3fd42518 and 41a1decf.