Re: [lmi] Transient ccache error

[Vadim Zeitlin]

On Wed, 22 Mar 2023 12:51:59 +0000 Greg Chicares <gchicares@sbcglobal.net> 
wrote:

GC> To debug this, I did 'make clean' and set "CCACHE_DEBUG=1",
GC> then rebuilt. Again, I saw numerous cache misses. I looked
GC> at some of the "*.ccache-log" files for x86-pc-gnu, and saw:
GC>   Compiler not expected to produce an object file
GC> However, googling for that yields:
GC>   No results found for "Compiler not expected to produce an object file"
GC> so this problem would seem to be quite rare.

 This messages comes from

https://github.com/ccache/ccache/blob/83a7d891dce9d0185be4d65306dd298dcd1620bb/src/ccache.cpp#L1173-L1176

which means that ctx.args_info.expect_output_obj was false in the run which
produced it which in turn means that -fsyntax-only (or, to be pedantic,
"-Zs", but this shouldn't be possible in our case) was used, as the only
way to set this field to false that I can see is here:

https://github.com/ccache/ccache/blob/83a7d891dce9d0185be4d65306dd298dcd1620bb/src/argprocessing.cpp#L728-L729

and it's set to true by default here:

https://github.com/ccache/ccache/blob/83a7d891dce9d0185be4d65306dd298dcd1620bb/src/ArgsInfo.hpp#L42


 And the only place where -fsyntax-only is used in lmi seems to be when
building %.physical_closure target, so it seems like it should somehow be
related to this, but I'm not sure how. I could investigate it further, but,
before I do, perhaps the connection between this error and the physical
closure target is already sufficient for you to understand what happens or,
at least, to reproduce it?

 If not, please let me know if you'd like me to continue with this.

 Thanks,
VZ

pgpBUa8i4ijy0.pgp
Description: PGP signature