Repository navigation
Description
A web project (Microsoft.NET.Sdk.Web) with CodoMetis.TypeKit.Generators 2.3.1 (Metalama 2026.1) does not build when SDK 11 is the only SDK in its .NET installation:
error LAMA0625: Analyzer 'Microsoft.CodeAnalysis.Razor.Compiler.dll' from .NET SDK 11.0.100-rc.3.26522.148 requires Roslyn 5.12.0.0, newer than the Roslyn 5.9.0.0 bundled with Metalama Compiler. Metalama searched all .NET SDKs installed on this machine for a compatible older version of 'Microsoft.CodeAnalysis.Razor.Compiler.dll', but none qualifies. [...] Installed SDKs found: 11.0.100.
Measured 2026-10-31 in spikes/Sdk11Rc3Metalama (the RC3 part of #170), on macOS arm64:
| Installation | 2.3.1 (Metalama 2026.1.30) | + Metalama.Framework 2026.1.31 |
+ Metalama.Framework 2027.0.5-preview |
|---|---|---|---|
SDK 11 RC3 alone, net11.0 |
fails, LAMA0625 |
fails, LAMA0625 |
builds |
SDK 11 RC3 alone, net10.0 |
fails, LAMA0625 |
not run | builds |
SDK 11 RC1 alone, net11.0 |
fails, LAMA0625 |
not run | builds |
| SDK 11 RC3 with SDK 10.0.403 in the same installation | builds, two LAMA0617 warnings |
not run | builds |
SDK 10.0.403, net10.0 |
builds | not run | builds |
A console project builds in every cell. So this is not an RC3 change: decision 120's RC1 runs had SDK 10.0.403 beside RC1, so they saw only LAMA0617, which the decision describes as "a warning, falling back to the SDK 10 Razor compiler". That fallback exists only when an SDK 10 is in the same installation; an SDK 11 container image or a CI job that installs SDK 11 alone has none.
Who it reaches. Every 2.x web application whose build moves to SDK 11 alone, whatever it targets, until it is on Metalama 2027 (2.5.0, decision 120) or references Metalama.Framework 2027 directly. It is loud (an error naming its fix), not silent.
Decided 2026-10-31 by decision 129: 2.4.0 moves the Generators package to Metalama 2027.0 (#173) and is tagged only on a stable 2027.0, with a CI lane that has only SDK 11 installed (#281). This issue closes when 2.4.0 ships. For 2.3.x, the way out on a machine with only SDK 11 is SDK 10 installed beside it, or a direct Metalama.Framework 2027 reference beside the 2.3.x packages; 2.4.0's changelog names both.
Evidence: spikes/Sdk11Rc3Metalama/README.md and KB-060, in PR #277.
Activity
CaffeinatedCoder commentedon Oct 31, 2026
Decided: decision 129 moves Metalama 2027.0 into 2.4.0 (#173), so 2.4.0 builds where SDK 11 is the only SDK. 2.4.0 is tagged only on a stable Metalama 2027.0; #281 adds the SDK-11-only CI lane that would have caught this. Closes when 2.4.0 ships.
CaffeinatedCoder commentedon Nov 11, 2026
State on 2026-11-11: .NET 11 is generally available, so SDK-11-only build images now exist outside previews. Metalama 2026.1.32 (2026-11-10) keeps
Metalama.Compiler2026.1.21, so its Roslyn is unchanged and this failure is expected to stand on SDK 11.0.100. That is read from the packages, not measured: the web rows above are to be repeated on the GA SDK with 2026.1.32 (#170). #284 adds a way out before the stable 2.4.0: a2.4.0-previewon a Metalama 2027 preview, if the licence-free build holds.