Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #1334 +/- ##
==========================================
+ Coverage 78.02% 79.53% +1.50%
==========================================
Files 61 61
Lines 4943 4945 +2
==========================================
+ Hits 3857 3933 +76
+ Misses 1086 1012 -74 🚀 New features to boost your workflow:
|
|
My interpretation of the test failure for macOS is that the orb package is not available there. |
|
Anything for which we provide binaries, we provide for both macOS and Linux equally. Hence if So whatever is wrong in CI, it's not because |
| @test ! GAP.Packages.load(path) | ||
|
|
||
| # Run package tests. | ||
| @test ! GAP.Packages.test("no_such_package") |
There was a problem hiding this comment.
Actually, I'd expect this to throw an exception (the input arguments were invalid). The values true / false should be reserved for "package was loaded, teste were run and passed / failed"
| # Make sure that GAP stores package information. | ||
| @test GAP.Packages.load(pkg[:name]) | ||
| # Manipulate GAP's global information such that | ||
| # `GAP.Globals.TestPackageAvailability` believes | ||
| # the package is not yet loaded. | ||
| # (Otherwise `GAP.Packages.build` would do nothing.) | ||
| pkg[:pkgloaded] = getproperty(GAP.Globals.GAPInfo.PackagesLoaded, pkg[:name]) | ||
| GAP.Wrappers.UNB_REC(GAP.Globals.GAPInfo.PackagesLoaded, GAP.RNamObj(String(pkg[:name]))) | ||
| pkg[:pkginfo] = collect(GAP.Globals.PackageInfo(GapObj(pkg[:name]))) | ||
| pkg[:avail_test] = [x.AvailabilityTest for x in pkg[:pkginfo]] |
There was a problem hiding this comment.
so we first load the package and then go through great lengths to convince GAP that it was not actually loaded in the first place. Why do we do that? The explanation given is
Make sure that GAP stores package information
But what does that mean? As far as I can tell, GAP.Globals.GAPInfo.PackagesInfo.orb is also populated with out that.
|
Ah -- not all macOS test fails. Only the "CI with GAP" tests. Those test GAP.jl against other GAP versions (namely GAP |
2 participants
addresses #1330
The nontrivial bit is to call
GAP.Packages.buildin a way that the function does something.The package in question must claim to be not available, thus one cannot simple take a loaded package or a package that has been installed with
GAP.Packages.install. The idea is to modify the information in the info records temporarily such that GAP thinks the package is not available.(This must be done package by package, not once before a loop, because PackageManager calls
InitializePackagesInfoRecords, which overwrites the changes in the info records.)