Re: Libidn 1 config glitch → determines ‘--no-install’ unsupported, uses it anyway

[Gordon Steemson]

Yes, those are exactly the ones.  Talking to the libtool people is probably a 
good idea.  Although there is a massive installed base of older libtools, I'd 
expect it can only be of help moving forward.

I'll try to figure out where to send such a request and get on that.

Gordon S.

> On Nov 20, 2024, at 12:55 AM, Simon Josefsson <simon@josefsson.org> wrote:
> 
> Can you show the warning messages you get?  Is it these warnings?
> 
> libtool: warning: '-no-install' is ignored for aarch64-apple-darwin23.6.0
> libtool: warning: assuming '-no-fast-install' instead
> 
> From a macOS build at https://gitlab.com/libidn/libidn/-/jobs/8321237800
> 
> Maybe we can ask the libtool people to lower this warning, I don't think
> it warns for anything that is actionable which makes it less useful.
> 
> /Simon
> 
> Gordon Steemson <gsteemso@gmail.com> writes:
> 
>> Building from tarballs, but that's not something that would change anything.
>> 
>> When you run "make check", the Makefile instructs libtool to do
>> various things.  Some of these involve a command line that includes a
>> "--no-install" flag, though whether this is passed to libtool or is
>> something libtool tells the linker or both I do not know.  As I
>> understand matters, whatever receives that flag does not understand it
>> when it is running on a Mac.  This causes the error channel to receive
>> a bright red "[warning]" message, with accompanying text to the effect
>> that "--no-install" is not understood – and, often but not always,
>> that "--no-fast-install" is being assumed instead.  This has no actual
>> effect that I can tell, but does clutter up the error log with dozens
>> of useless lines.  (Some other packages produce over a hundred of
>> them, so it could be worse.)
>> 
>> Gordon S.
>> 
>>>> On Nov 19, 2024, at 2:21 PM, Simon Josefsson <simon@josefsson.org> wrote:
>>> 
>>> What is the actual problem you are seeing?  Are you building from
>>> tarballs, or from git?
>>> 
>>> /Simon
>>> 
>>> Gordon Steemson <gsteemso@gmail.com> writes:
>>> 
>>>> Ah, my apologies.  "--no-install" is not an argument someone might
>>>> give to configure, it is something Libtool uses (most often seen
>>>> during "make check"), and I was proposing that a configure test should
>>>> exist to determine when it shouldn't do so.
>>>> 
>>>> I don't know whether such a thing has already been written or if
>>>> various other projects' Makefiles simply never pass "--no-install" to
>>>> libtool in the first place.  Either way, libidn is far from the only
>>>> package that DOES use it, and it's more a cosmetic annoyance than
>>>> anything serious, unless the person installing the software is
>>>> particularly twitchy about warning messages.
>>>> 
>>>> Gordon S.
>>>> 
>>>>>> On Nov 19, 2024, at 11:15 AM, Simon Josefsson <simon@josefsson.org> 
>>>>>> wrote:
>>>>> 
>>>>> What did you expect ./configure --no-install to do?  I don't think it is
>>>>> a common ./configure parameter with any well-established semantics.  How
>>>>> does 'make check' use --no-install?  I don't know what '--no-install'
>>>>> refers to really.
>>>>> 
>>>>> /Simon
>>>>> 
>>>>> Gordon Steemson <gsteemso@gmail.com> writes:
>>>>> 
>>>>>> Um.  False alarm.  ./configure does not actually check for 
>>>>>> ‘--no-install’ – I had it mixed up with another test that it _does_ 
>>>>>> perform.
>>>>>> 
>>>>>> (Maybe it should test for that in future?)
>>>>>> 
>>>>>> Gordon S.
>>>>>> 
>>>>>>>> On Nov 16, 2024, at 9:18 PM, Gordon Steemson <gsteemso@gmail.com> 
>>>>>>>> wrote:
>>>>>>> 
>>>>>>> What the subject line says.  I built the most recent libidn 1
>>>>>>> package, v. 1.42, and even though ./configure correctly observed
>>>>>>> that ‘--no-install’ is not understood by my Mac, `make check` used
>>>>>>> it liberally anyway.
>>>>>>> 
>>>>>>> I realize this is very far from an Earth-shattering problem, but
>>>>>>> seeing that much warning-label red lettering go past on the console
>>>>>>> is not a fun experience in the time before you work out that it’s
>>>>>>> not anything _important_.
>>>>>>> 
>>>>>>> Sincerely,
>>>>>>> Gordon Steemson
>>>>>> 
>>>>>> 
>>>>> <signature.asc>
>>>> 
>>>> 
>>> <signature.asc>
> <signature.asc>