Having a compiler set in a depends_on directive enforces that compiler by default only on that node, while the preferred compiler is used for other nodes.

Steps to reproduce the issue

Modify hdf5 with this diff:

--- a/var/spack/repos/builtin/packages/hdf5/package.py
+++ b/var/spack/repos/builtin/packages/hdf5/package.py
@@ -83,7 +83,7 @@ class Hdf5(AutotoolsPackage):
     if sys.platform != 'darwin':
         depends_on('numactl', when='+mpi+fortran')
     depends_on('szip', when='+szip')
-    depends_on('[email protected]:')
+    depends_on('[email protected]: %[email protected]')
 
     # There are several officially unsupported combinations of the features:
     # 1. Thread safety is not guaranteed via high-level C-API but in some cases

then try the following:

$ spack compiler list
==> Available compilers
-- clang ubuntu18.04-x86_64 -------------------------------------
[email protected]  [email protected]  [email protected]  [email protected]  [email protected]  [email protected]  [email protected]

-- gcc ubuntu18.04-x86_64 ---------------------------------------
[email protected]  [email protected]  [email protected]  [email protected]  [email protected]  [email protected]  [email protected]

$ spack solve hdf5~mpi
==> Best of 0 answers.
==> Optimization: [0, 2, -11, 0, 0, 0, -3, 1, 0, 1, -1, 0]
[email protected]%[email protected]~cxx~debug~fortran~hl~java~mpi+pic+shared~szip~threadsafe api=v18 arch=linux-ubuntu18.04-x86_64
    ^[email protected]%[email protected]+optimize+pic+shared arch=linux-ubuntu18.04-x86_64

Error Message

No error message, but one would expect to see by default:

$ spack solve hdf5~mpi
==> Best of 0 answers.
==> Optimization: [0, 2, -11, 0, 0, 0, -3, 1, 0, 1, -1, 0]
[email protected]%[email protected]~cxx~debug~fortran~hl~java~mpi+pic+shared~szip~threadsafe api=v18 arch=linux-ubuntu18.04-x86_64
    ^[email protected]%[email protected]+optimize+pic+shared arch=linux-ubuntu18.04-x86_64

Information on your system

• Spack: 0.16.0-70-44665cb4e6
• Python: 3.8.6
• Platform: linux-ubuntu18.04-broadwell
• Concretizer: clingo

Additional information

It may seem silly to enforce a compiler in a depends_on directive, but this is what happens under the hood when a spec containing a compiler specification is concretized together with others in an environment. For instance:

spack:
  view: false
  concretization: together
  config:
    concretizer: clingo
  specs:
  - hdf5%[email protected] ^zlib

behaves differently if concretization: separately is used.

• I have run spack debug report and reported the version of Spack/Python/Platform
• I have searched the issues of this repo and believe this is not a duplicate
• I have run the failing commands in debug mode and reported the output