Re: Attempting to convert -CURRENT host to pkgbase, fails on none-existent kernel config.

On 2026-08-23 04:47, Mark Millard wrote:
> On 8/22/26 03:44, Alastair Hogge wrote:
>> On 2026-08-14 19:34, Lexi Winter wrote:
>>> Alastair Hogge wrote in <1be98630bb8ef5f3cc1c665ca2f3961d@riseup.net>:
>>>> I am attempting to convert my build host/NAS to pkgbase. In the host's
>>>> make.conf I have:
>>>> KERNCONF= FAFNIR FAFNIR-DEBUG GENERIC GENERIC-NODEBUG
>>>> During the "make packages" phase, it fails because it cannot find
>>>> DIREWOLF_DEBUG in /sys/amd64/conf.
>>>
>>> what package is looking for this file?  is it possible that you deleted
>>> this file on disk but didn't remove it from Git?  in that case, src-sys
>>> will fail as it uses the Git index to determine what files to package.
>> 
>> You are probably right about that, I nuked ${SRC} and re-applied local
>> diffs and that error is gone.
>>  
>>> if that's not the issue, please make available the full output from
>>> buildkernel, buildworld and (update-)packages.
>> 
>> https://bugs.freebsd.org/bugzilla//show_bug.cgi?id=297183
>> 
>> 
>> 
> 
> bugzilla was having backend fetch problems preventing subnimttal of a
> comment, so I report here:
> 
> 
> From what I see the fatal status is just a consequence of having
> DEVELOPER_MODE enabled:
> 
> ctx.developer_mode = pkg_object_bool(pkg_config_get("DEVELOPER_MODE"));
> 
> and:
> 
> libpkg/pkg.c
> 
> . . .
> int
> pkg_addfile_attr(struct pkg *pkg, const char *path, const char *sum,
> 		 const char *uname, const char *gname, mode_t perm,
> 		 u_long fflags, time_t mtime,
> 		 const char *symlink_target, bool check_duplicates)
> {
> . . .
> 	struct pkg_file *existing = pkg_filev_insert_sorted(&pkg->files, f);
> 	if (existing != NULL) {
> 		pkg_file_free_content(&f);
> 		if (check_duplicates) {
> 			if (ctx.developer_mode) {
> 				pkg_emit_error("duplicate file listing: %s, fatal (developer mode)",
> path);
> 				return (EPKG_FATAL);
> 			} else {
> 				pkg_emit_error("duplicate file listing: %s, ignoring", path);
> 			}
> 		}
> 		return (EPKG_OK);
> 	}
> 
> 	return (EPKG_OK);
> . . .
> 
> 
> If you can operate with DEVELOPER_MODE disabled, the specific
> type of stopping operation behavior should not occur.

Ah yes, you are correct. I have DEVELOPER_MODE defined in
${HOME}/.profile. A simple 'unset' at the top of my POSIX build scripts
have resolved the issue.

Thanks.