Re: [Tinycc-devel] Differences between libtcc in "0.9.27 (x86_64 Linux)" and "0.9.27 - 0378168 (x86_64 Linux)" and some unexpected behavior

[Dwight Schauer]

On Thu, 2021-06-10 at 20:01 +0200, grischka wrote:
> Dwight Schauer wrote:
> > With "0.9.27 - 0378168" I get some unexpected behavior.
> > 
> > 1) the byte sizes of compilation units (separate libtcc states)
> > are about 8K larger each than with the stock libtcc
> 
> runtime memory now gets permissions (r/w/x) set per section more
> accurately however this means that sections need to start on a new
> memory page each (of size 4096 usually).
> 
> Can be disabled with CONFIG_RUNMEM_RO in tccrun,c.

Thanks, understood.

> > 2) with "-Wunsupported -Wall -Werror" tcc_compile_string exits the
> > entire program rather than returning non zero for non supported or
> > bogus flags.
> 
> ???. tcc_compile_string() does not take any flags.
> 
> It is true that tcc_set_options() would call exit() when it should
> not
> which is not a new problem however.

Yeah, I meant by calling tcc_set_options() before calling
tcc_compile_string(). Understood.

> 
> > 3) compilation errors don't result in a non zero exit
> 
> I can not reproduce this, in that generality.
> 

I'd need to create a reproduction sample...

> > 4) "-b" needs __bound_strlen, but the new libtcc.so does not have
> > that
> > symbol
> 
> In fact libtcc1.a doesn't contain bcheck.o anymore.  Therefor with
> -b in combination with -nostdlib, bcheck.o needs to be added
> explicitly.

Thanks.