NetBSD-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Building NetBSD 11 Release from Source Fails on Linux/MacOS



Super cool! Thank you for identifying the cause of this
issue!

I just applied the patch from https://gnats.netbsd.org/61011
and everything builds as expected. I love it.

Indeed, on the three machines on which the build failed, the
path to the NetBSD src tree contain a `C` component. They all
have that in common and I did not think about it as a
difference to the VMs.

I am sorry for making it more complicated than necessary. In
my initial emails, I shortened paths in the hope that they
were less confusing when being shorter. In the later
messages, I did so since they were all slightly different
over the various virtual and physical machines (except of the
three ones where the build failed). I keep for the future
that I should not do that.

Some questions for me to learn: 

1. How precisely did you call strace? 
   I assume that you took the failing commands from the logs 
   and straced it. But which arguments and switches do you 
   use then?
2. Did you directly trace for file reads or did you just 
   trace all system calls?
3. Once you have all the output from strace logged, for what 
   do you search in case of this issue?  
   Just anything where an .m4 file appears?

Thank you once more for identifying and fixing this issue!


Best,
Helge 

On Sat, 2026-11-14 at 22:39 +0000, RVP wrote:
> On Fri, 13 Nov 2026, Helge Pfeiffer wrote:
> 
> > I am sending the complete bundle again, to be sure to
> > have
> > the correct log files for the obj dir per build. Since I
> > already deleted everything on MacOS Sequoia after my last
> > message, I recreated everything for the build on it. That
> > is,
> > it is likely best to disregard the link from my last
> > message
> > and just use the link above.
> > 
> 
> Since I could see the actual paths involved now, figuring
> out what was wrong
> wasn't difficult--after running strace(1) on the m4
> command-line, and perusing
> libgmp's .m4 files.
> 
> Short version: you have a directory named `C' (yes,
> really!). The long
> version is in PR toolchain/61011.
> 
> -RVP



Home | Main Index | Thread Index | Old Index