[syslinux] boot... round 2

[Adam Williamson]

On Fri, 2015-07-03 at 22:39 +0200, Thomas Schmitt wrote:
> Hi,
> 
> Adam Williamson
> > I think it would be better if someone else confirms there's 
> > actually
> > something to fix.
> 
> There was indeed the shortcomming that byte values 128 to 255
> were treated as whitespace and mapped to "_" resp. "".
> This would affect non-ASCII characters of UTF-8 or old
> ISO-8859.
> 
> 
> But the reason why gcc 5 made a difference to gcc 4.9
> is not found yet.
> The question why poma's observations are not reproducible
> for you needs an answer, too.
> 
> Your observations could match e.g. a trailing garbage byte
> which gcc 5 causes in SYSAPPEND strings and which gets dropped
> by the now functional whitespace converter. Hopefully some byte
> in the range of 1 to 31.
> 
> poma's observations would be explainable if on his system
> a byte in the range of 128 to 255 in one of the SYSAPPEND
> strings has to survive. (But why did "unsigned char c;"
> not solve the problem for poma ? Why did it work with gcc 4.9 ?)
> 
> One should create a possibility to dump the SYSAPPEND
> strings. There is a function
>   ./core/sysappend.c:void print_sysappend(void)
> but i cannot spot any occasion where it gets called.
> 
> 
> Have a nice day :)

I suppose it's possible we were testing with different syslinux
configurations.  I was building Fedora Rawhide live images as my test,
with the volume label '20150702-syslinux-x86_64'. The volume label
chosen for live images does in fact appear in the syslinux menu text,
for Fedora live images.

I don't know what kind of test poma was using.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net