Re: [Tinycc-devel] Large files causes infinite loop on mac

[Herman ten Brugge]

On 12/27/22 07:58, Herman ten Brugge wrote:
> On 12/27/22 05:45, Levo D wrote:
> > I'm messing around with tcc. I don't need this to be fixed, I just 
> > thought someone may want the report. I'm trying to figure out how 
> > fast the tcc backend is for my compiler.
> > I seem to hit an infinite loop on mac osx venture. I built tcc 
> > earlier this month (Dec 2022). I suspect it's an overflow bug. Here's 
> > how to reproduce, on linux x86-64 it errors out at a smaller number.
> >
> > Run the following python script (I copied it below) 
> > https://bolinlang.com/genc.py
> >
> > python3 ./genc.py 2200000 4
> >
> > Create object files then try to link. It'll loop forever
> >
> > tcc -c *.c
> > tcc *.o
> >
> > You can try building it straight which will build but won't execute
> >
> > $ tcc *.c
> > $ ./a.out
> > dyld[4330]: dyld cache '(null)' not loaded: syscall to map cache into 
> > shared region failed
> > dyld[4330]: Library not loaded: /usr/lib/libSystem.B.dylib
> >    Referenced from: <no uuid> /private/tmp/t/test/a.out
> >    Reason: tried: '/usr/lib/libSystem.B.dylib' (no such file), 
> > '/System/Volumes/Preboot/Cryptexes/OS/usr/lib/libSystem.B.dylib' (no 
> > such file), '/usr/lib/libSystem.B.dylib' (no such file, no dyld 
> > cache), '/usr/local/lib/libSystem.B.dylib' (no such file)
> > Abort trap: 6
> >
> > Using a smaller number everything works fine. I'll repeat that I'm 
> > just playing around and don't need this to work. Here's a copy paste 
> > of the python script
> I fixed the forever loop.
> The code now compiles but does not run. I get:
>
> dyld[74088]: dyld cache 
> '/System/Library/dyld/dyld_shared_cache_arm64e' not loaded: syscall to 
> map cache into shared region failed
> dyld[74088]: Library not loaded: '/usr/lib/libSystem.B.dylib'
>   Referenced from: '/Users/hermantb/tinycc/tst/a.out'
>   Reason: tried: '/usr/lib/libSystem.B.dylib' (no such file), 
> '/usr/local/lib/libSystem.B.dylib' (no such file)
> Abort trap: 6
>
> Probably because the bss section is larger then 4G.
>
> I also tried to compile you code with gcc/clang on fedora(x86_64) and 
> also get overflow reports when compiling.
> So probably there is a limit some where. I did not check.
>
> So if gcc/clang cannot compile the code tcc can fail as well.
I updated the python script. I can now run:

$ python3 ./genc.py 100000000 100
$ tcc -bench -c *.c
* 78156 idents, 1157015 lines, 28090333 bytes
* 91.295 s, 12673 lines/s, 0.3 MB/s
* text 11784304, data.rw 0, data.ro 0, bss 4000000 bytes
$ tcc -bench *.o
* 0 idents, 0 lines, 1 bytes
* 3.383 s, 0 lines/s, 0.0 MB/s
* text 0, data.rw 0, data.ro 0, bss 0 bytes

updated lines in script:
29c29
<                 file.write(f"\tstatic int array[{nextFunctionIndex}];\n")
---
>                 file.write(f"\tstatic int 
array[{nextFunctionIndex-functionStartIndex}];\n")
33c33
<                 r = math.floor(random.random()*nextFunctionIndex)
---
>                 r = 
math.floor(random.random()*(nextFunctionIndex-functionStartIndex))