Summary: | adapt x11/cool-retro-term for clang-on-powerpc | ||||||
---|---|---|---|---|---|---|---|
Product: | Ports & Packages | Reporter: | Mark Linimon <linimon> | ||||
Component: | Individual Port(s) | Assignee: | Mark Linimon <linimon> | ||||
Status: | Closed FIXED | ||||||
Severity: | Affects Only Me | CC: | danfe | ||||
Priority: | --- | Flags: | bugzilla:
maintainer-feedback?
(danfe) |
||||
Version: | Latest | ||||||
Hardware: | Any | ||||||
OS: | Any | ||||||
Attachments: |
|
A commit references this bug: Author: linimon Date: Fri Oct 4 22:40:29 UTC 2019 New revision: 513793 URL: https://svnweb.freebsd.org/changeset/ports/513793 Log: Prepare for powerpc-on-clang by deleting hard-coded tests for architecture as a stand-in for "are we running on gcc". (PR 239181) As a side-effect, adding the stanza MAKE_ENV= LLVM_CONFIG=${LOCALBASE}/bin/llvm-config${LLVM_DEFAULT} obviates the need for hardcoding definitions for CC, CXX, and CPP (later linimon testing). PR: 239181 Approved by: danfe (maintainer-timeout, > 2 months) Changes: head/x11/cool-retro-term/Makefile Sorry, this PR somehow went off my radar. I don't quite understand the patch, especially why did you have to pessimize most common case of clang-based tier-1 arches by pulling devel/llvmXX from ports instead of using system compiler (like it was before). No need to take any action, I'll fix this mess soonish. (In reply to Alexey Dokuchaev from comment #2) There is currently no way to tell e.g. Mk/Uses/compiler.mk "use clang from base if available, otherwise use clang from ports". This would be the correct fix here. Patches welcome. Hardcoding CC etc. is always wrong: a) it overrides possible user changes for testing; b) it issues a useless warning message on GCC-based systems. Please see https://wiki.freebsd.org/HardcodedLLVMVersions#Ports_That_Redefine_CC_.2F_CPP_.2F_CXX . > "use clang from base if available, otherwise use clang from ports"
Ah, OK, now I see why you got carried away in that direction. :-)
No, AFAIR it never needed specifically Clang. Those "FOO=clang" were remnants from the early 9.X-ish times to avoid pulling external compiler, and Clang just hit the base system but was not named cc/c++ yet. This is ugly and should had been fixed long ago.
I suspect something simple like compiler:c++11-lang would suffice here, in case patching the source code to make it gcc-4.2.1 friendly would not be feasible.
A commit references this bug: Author: danfe Date: Wed Oct 23 12:11:24 UTC 2019 New revision: 515252 URL: https://svnweb.freebsd.org/changeset/ports/515252 Log: - Switch to USES+=compiler:c++11-lang to unbreak on PowerPC instead of the approach taken in r513793 (Clang was never required for this port, any not-so-ancient compiler would suffice) - Sort the values on the USES list; avoid hardcoding version numbers - Pluralize the noun in the COMMENT line and make it a bit shorter Reviewed by: pkubaj PR: 239181 Changes: head/x11/cool-retro-term/Makefile |
Created attachment 205742 [details] patch to x11/cool-retro-term/Makefile Prepare for powerpc-on-clang by deleting hard-coded tests for architecture as a stand-in for "are we running on gcc".