cluts daily reports 08/11 - fixes to buf and alloc

cluts daily reports 08/11 - fixes to buf and alloc

[Luka Marčetić]

[-- Attachment #1: Type: text/plain, Size: 189 bytes --]

Buf should work now I guess (comitted), alloc still needs work around 
for some specifics - albeit arguably rational - of musl.
So that's the priority, along with finishing pthread_eintr.


[-- Attachment #2: to-report --]
[-- Type: text/plain, Size: 151 bytes --]

 tests/alloc.c                                                     |   35 
 tests/buf.c                                                       |  113 -

Re: cluts daily reports 08/11 - fixes to buf and alloc

[Solar Designer]

Luka -

On Thu, Aug 11, 2011 at 04:14:08AM +0200, Luka Marčetić wrote:
> Buf should work now I guess (comitted), alloc still needs work around 
> for some specifics - albeit arguably rational - of musl.
> So that's the priority, along with finishing pthread_eintr.
> 
>  tests/alloc.c                                                     |   35 
>  tests/buf.c                                                       |  113 -

Thanks for the report.  Where's the final line of diffstat output,
though?  The above says that you mostly deleted a few lines from buf.c,
but it does not tell how many and whether you also added/edited
anything.  Overall, it does show that you made very little progress.

You should try not only to fix problems in code already written, but
also make progress at further tasks each and every day.  Maybe limit
your fixing of existing code to 4 hours per day (this sounds excessive
to me, but it works as an upper limit), and use the rest of time
(4+ hours per day since you need to work hard during the remaining GSoC
time) for writing new code.  How does this sound to you?

Alexander

Re: cluts daily reports 08/11 - fixes to buf and alloc

[Luka Marčetić]

On 08/11/2011 09:08 PM, Solar Designer wrote:
> Luka -
>
> On Thu, Aug 11, 2011 at 04:14:08AM +0200, Luka Marčetić wrote:
>> Buf should work now I guess (comitted), alloc still needs work around
>> for some specifics - albeit arguably rational - of musl.
>> So that's the priority, along with finishing pthread_eintr.
>>
>>   tests/alloc.c                                                     |   35
>>   tests/buf.c                                                       |  113 -
> Thanks for the report.  Where's the final line of diffstat output,
> though?  The above says that you mostly deleted a few lines from buf.c,
> but it does not tell how many and whether you also added/edited
> anything.  Overall, it does show that you made very little progress.

Hey, I just checked my e-mail. I removed ~ files, binaries, git data 
etc, which is why I thought the last line made little sense.

> You should try not only to fix problems in code already written, but
> also make progress at further tasks each and every day.

Buggy code doesn't help much. That said, I did focus on pthread_eintr 
today, rather than alloc. I'm pretty stuck with alloc. I just can't get 
it to work with musl. Valgrind is clean, strace says little (or perhaps 
I just can't interpret it well), and gdb - it just won't follow the 
second child, so I can't step through it as I hoped to do.

>   Maybe limit
> your fixing of existing code to 4 hours per day (this sounds excessive
> to me, but it works as an upper limit), and use the rest of time
> (4+ hours per day since you need to work hard during the remaining GSoC
> time) for writing new code.  How does this sound to you?

The thing is, writing code means writing bugs, so it's just a matter of 
time when the fine line is crossed. However, I'll try to take on other 
tasks as well - esp. those that Rich suggested.
Luka

Re: cluts daily reports 08/11 - fixes to buf and alloc

[Solar Designer]

On Fri, Aug 12, 2011 at 04:21:59AM +0200, Luka Mar??eti?? wrote:
> On 08/11/2011 09:08 PM, Solar Designer wrote:
> >Thanks for the report.  Where's the final line of diffstat output,
> >though?  The above says that you mostly deleted a few lines from buf.c,
> >but it does not tell how many and whether you also added/edited
> >anything.  Overall, it does show that you made very little progress.
> 
> Hey, I just checked my e-mail. I removed ~ files, binaries, git data 
> etc, which is why I thought the last line made little sense.

Going forward, please remove those extra files before you run
diff|diffstat.  Don't edit diffstat output.  If needed, you may make
copies of the cluts tree and remove extra files from the copies only.

> >You should try not only to fix problems in code already written, but
> >also make progress at further tasks each and every day.
> 
> Buggy code doesn't help much.

True, but you can't afford to spend over 50% of the remaining GSoC time
on fixing bugs and getting stuck on some of them.  This would be OK if
you already had everything implemented, but you don't.

> That said, I did focus on pthread_eintr today, rather than alloc.

OK.

> I'm pretty stuck with alloc. I just can't get 
> it to work with musl. Valgrind is clean, strace says little (or perhaps 
> I just can't interpret it well), and gdb - it just won't follow the 
> second child, so I can't step through it as I hoped to do.

This may or may not be the best approach, but you may try inserting:

asm("int3");

just in the child process or thread that you want interrupted.  You may
even make the int3 conditional in the code (check whatever condition you
like with an if).  Or at high level, you may set the same breakpoint
from gdb itself.

Alexander