[Toybox] [landley/toybox] xargs: exec ... Argument list too long (#40)

[Rob Landley]

And after deleting the mailing list archive again, dreamhost restored
from a 3 day old backup that doesn't contain this entire thread.

Wheee...

Rob

On 09/28/2017 10:18 AM, enh wrote:
> Did you look at my xargs patch in this thread? It handles all that. I
> think the find patch is pretty similar, I just haven't had time to set
> up a working test case. (Or rather I fell into ratholes while doing so.)
> 
> On Sep 28, 2017 04:41, "Rob Landley" <rob at landley.net
> <mailto:rob at landley.net>> wrote:
> 
>     Sorry, yes I did get time to look at this, I just haven't _fixed_ it
>     yet.
> 
>     The problem is that environment variables and arguments both fit in the
>     same space, which is dynamic at 1/4 the stack size limit as mentioned at
>     http://lists.landley.net/pipermail/toybox-landley.net/2016-August/008592.html
>     <http://lists.landley.net/pipermail/toybox-landley.net/2016-August/008592.html>
>     last time around.
> 
>     Which means that _any_ static size we enforce isn't guaranteed to work,
>     if you have a lot of environment variables defined (or just one really
>     big one). Environment varaibles can even consume the space to the point
>     you can't launch a _single_ argument. (If it's less than PATH_MAX, we
>     have a problem.)
> 
>     So I think it has to probe, and be dynamic, and at least not _loop_ when
>     find produces a single long name that's more than the remaining
>     environment space to launch a child. (I.E. test the corner case that the
>     maximum number of args we can fit is zero.)
> 
>     Space consumption should be exactly sizeof(char *)*(argc+1) +
>     sizeof(char *)+(envc+1) + the strlen+1 of all the argv[] + the strlen+1
>     of all the envp[]. I need to test that's still the case (it used to be
>     when the limit was hardwired 32*4096).
> 
>     Sorry for the delay, still juggling way more things than I have energy
>     for. :(
> 
>     Rob
>