ARG_MAX missing from getconf

ARG_MAX missing from getconf

[Matthew "mirage335" Hines]

[-- Attachment #1: Type: text/plain, Size: 265 bytes --]

Really hope other Cygwin packages will not be broken because 'ARG_MAX' is
now undefined in 'getconf' output. Caught this in a CI test for a rather
large shell script.

https://github.com/mirage335-colossus/ubiquitous_bash/blob/master/structure/installation.sh#L385

Re: ARG_MAX missing from getconf

[Corinna Vinschen]

On Sep  8 00:14, Matthew "mirage335" Hines via Cygwin wrote:
> Really hope other Cygwin packages will not be broken because 'ARG_MAX' is
> now undefined in 'getconf' output. Caught this in a CI test for a rather
> large shell script.
> 
> https://github.com/mirage335-colossus/ubiquitous_bash/blob/master/structure/installation.sh#L385

Please read this, this is supposed to explain the new ARG_MAX
behaviour sufficiently:

https://cygwin.com/pipermail/cygwin/2023-August/254352.html


Corinna

Re: ARG_MAX missing from getconf

[Matthew "mirage335" Hines]

[-- Attachment #1: Type: text/plain, Size: 1254 bytes --]

Ok, I guess that makes sense.

My worry was some './configure' script determining that the build
environment is somehow not 'sane'. At least for the limited set of programs
that I use under cygwin, I have set up my CI to now watch for changes in
which binaries are present. Nothing obviously significant in that history
so far.

At this point, from comparing with my previous CI results, I am guessing
this doesn't affect me, aside from adding an exception to my test case for
this specifically. Might be better to get closer to the usual Linux
behavior I am familiar with, of a much higher, valid, ARG_MAX value.

On Fri, Sep 8, 2023 at 6:49 AM Corinna Vinschen <corinna-cygwin@cygwin.com>
wrote:

> On Sep  8 00:14, Matthew "mirage335" Hines via Cygwin wrote:
> > Really hope other Cygwin packages will not be broken because 'ARG_MAX' is
> > now undefined in 'getconf' output. Caught this in a CI test for a rather
> > large shell script.
> >
> >
> https://github.com/mirage335-colossus/ubiquitous_bash/blob/master/structure/installation.sh#L385
>
> Please read this, this is supposed to explain the new ARG_MAX
> behaviour sufficiently:
>
> https://cygwin.com/pipermail/cygwin/2023-August/254352.html
>
>
> Corinna
>

Re: ARG_MAX missing from getconf

[Corinna Vinschen]

On Sep  8 07:36, Matthew "mirage335" Hines via Cygwin wrote:
> Ok, I guess that makes sense.
> 
> My worry was some './configure' script determining that the build
> environment is somehow not 'sane'. At least for the limited set of programs
> that I use under cygwin, I have set up my CI to now watch for changes in
> which binaries are present. Nothing obviously significant in that history
> so far.
> 
> At this point, from comparing with my previous CI results, I am guessing
> this doesn't affect me, aside from adding an exception to my test case for
> this specifically. Might be better to get closer to the usual Linux
> behavior I am familiar with, of a much higher, valid, ARG_MAX value.

We just don't have a maximum except an out of resources situation.  A
standard-conformant application must deal with any sysconf variable
denoting a limit to return -1, i. e., "no limit".

Corinna

Re: ARG_MAX missing from getconf

[Matthew "mirage335" Hines]

[-- Attachment #1: Type: text/plain, Size: 1091 bytes --]

Ok, excellent, thank you, good to know.

On Fri, Sep 8, 2023 at 7:44 AM Corinna Vinschen <corinna-cygwin@cygwin.com>
wrote:

> On Sep  8 07:36, Matthew "mirage335" Hines via Cygwin wrote:
> > Ok, I guess that makes sense.
> >
> > My worry was some './configure' script determining that the build
> > environment is somehow not 'sane'. At least for the limited set of
> programs
> > that I use under cygwin, I have set up my CI to now watch for changes in
> > which binaries are present. Nothing obviously significant in that history
> > so far.
> >
> > At this point, from comparing with my previous CI results, I am guessing
> > this doesn't affect me, aside from adding an exception to my test case
> for
> > this specifically. Might be better to get closer to the usual Linux
> > behavior I am familiar with, of a much higher, valid, ARG_MAX value.
>
> We just don't have a maximum except an out of resources situation.  A
> standard-conformant application must deal with any sysconf variable
> denoting a limit to return -1, i. e., "no limit".
>
> Corinna
>

Re: ARG_MAX missing from getconf

[Takashi Yano]

On Fri, 8 Sep 2023 07:36:48 -0400
"Matthew \"mirage335\" Hines wrote:
> Ok, I guess that makes sense.
> 
> My worry was some './configure' script determining that the build
> environment is somehow not 'sane'. At least for the limited set of programs
> that I use under cygwin, I have set up my CI to now watch for changes in
> which binaries are present. Nothing obviously significant in that history
> so far.
> 
> At this point, from comparing with my previous CI results, I am guessing
> this doesn't affect me, aside from adding an exception to my test case for
> this specifically. Might be better to get closer to the usual Linux
> behavior I am familiar with, of a much higher, valid, ARG_MAX value.

For what purpose do you use ARG_MAX value?
In Linux, ARG_MAX value varies by "ulimit -s xxxx" command.
In my environment, it's 2097152 by default, however, after
ulimit -s 1024
, it will be reduced to 262144.

So, I think it is not good idea to use ARG_MAX value for the
purpose other than its original purpose.

-- 
Takashi Yano <takashi.yano@nifty.ne.jp>