Mail Archives: djgpp/2016/05/05/00:25:55
On 05/05/2016 04:04 AM, Wiktor S. (wswiktorSP AT Mpoczta DOT fm) [via djgpp AT delorie DOT com] wrote:
>> later versions installed from binaries are not OK. Try compiling
>> "Hello World" style program which uses <iostream> repeatedly under
>> WIn10 to get random failures. I'm getting them in average once per hundred or several hundreds
>> of compiles. In other
>> test I only repeatedly run preprocessor and compared it outputs.
>
>
> On a 150 MB FAT16 ramdisk I installed the following zips:
>
> bnu226br3
> gcc610b
> gpp610b
> djdev205
>
>
> On the first trial, it failed at 7th iteration with bogus compilation error inside iostreams.
>
> Consecutive trials seem to work longer and longer: 8, 14, 36, 84.
>
> So I restarted the machine. Again, failure on 5th attempt.
> That's nowhere near "several hundreds" you are getting.
I mentioned approximate frequency which I had. It could depend on different additional conditions
when some instability is present:
Some differences:
- I used Win 10 Home virtual machine under Linux
- no ramdrive
Also tested Win 10 Enterprise trial version earlier last autumn with the same results
One thing we should find to get actual GCC version where the problems actually start
1) gcc-4.4.7 works OK for me under Win 10. Files are available from
http://ap1.pp.fi/djgpp/gcc/4.4.7/v205/
Have not seen similar pre-processor failure under WIn10 in about half a year with it.
2) gcc-4.5.3 seems to work OK under WIn10, I rebuilt it under WIn10 yesterday, so rather
limited testing has been done. My test script has not got a single failure in more than 20000
pre-processor runs. Suspected also larger relocation count support in new binutils version.
I used optimisation level -Os, no debug information and binutils-2.20 (before support of large
relocation counts were added). Files are at:
http://ap1.pp.fi/djgpp/gcc/4.5.3/v205/
3) gcc-4.6.4: bootstrapped using gcc-4.4.7. I used optimisation level -Os, no debug information
and binutils-2.20 in last build. I git random several pre-processor failures in bootstrap stage3.
I pushed it through with force by repeating './djmake.sh bootstrap'. Final compiler fails in my
pre-processor torture test
Some observations:
- more than 64K relocations support in binutils (beginning from version 2.22) does not seem
to be related to this problem
- bootstrap comparison succeeds when the problem appears and I got bootstrap still to finish
by repeating it (it not always possible though).
About higher failure frequency mentioned above:
- it would be nice to try 4.4.7 and 4.5.3 on these systems.
Andris
- Raw text -