| Summary: | sysutils/grub2-pcbsd fails build if tigcc installed due to detecting gcc from tigcc as the gcc to use for portions of build | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Ports & Packages | Reporter: | alt2600 | ||||||
| Component: | Individual Port(s) | Assignee: | freebsd-ports-bugs (Nobody) <ports-bugs> | ||||||
| Status: | New --- | ||||||||
| Severity: | Affects Only Me | Flags: | bugzilla:
maintainer-feedback?
(kevlo) |
||||||
| Priority: | --- | ||||||||
| Version: | Latest | ||||||||
| Hardware: | amd64 | ||||||||
| OS: | Any | ||||||||
| Attachments: |
|
||||||||
|
Description
alt2600
2021-08-07 17:32:01 UTC
Created attachment 227010 [details]
configure log
Configure log
Created attachment 227011 [details]
build log
Build Log
so I explored this some more by building in a clean bhyve and my base system and comparing tee'd configure logs and config.log from work directory. Still exploring how to mitigate but updating bug for actual error, its possible I'm sharing environments with sudo and that is breaking things as root shouldn't have this in their path. I know I made this change on a linux box due to RHEL fun times, and assumed it was safe on freebsd and forgot about it, I might not have done this, it might be the default here, but i digress. For some reason if tigcc is installed, it is selected by configure for use building the port. it seems like TIGCC expects it is in the path, and this breaks things in wonderfully fun ways. I noticed between logs the endianess of the system said big endian on my amd64 which was obviously wrong, and this led to finding the tigcc issue. its not clear why configure would accept a 68k compiler even if it was found when not cross-compiling. It appears it was using gcc10 for the start and gcc (tigcc) for the build stage that crashed at gcc -o garbage-gen -DGRUB_FILE=\"util/garbage-gen.c\" -I. -I. -I. -I. -I./include -I./include -I./grub-core/lib/libgcrypt-grub/src/ util/garbage-gen.c As this issue is caused by placing tigcc into my users .shrc path, not the system path during install I will mark this resolved. not a bug, but allow others to maybe find in google if they see errors. I'll give a little time before doing so the maintainer sees and can let upstream know, if they think maybe it is a configure bug. They can mark resolved in that case if they want. I still think it is odd the port us mixing which compiler to use for each component though, or that it ever accepted a 68k compiler on an amd64 host in the first place. ^Triage: commit bit was taken in for safekeeping. |