llvm dev - Jul 2021 - Binary utilities: switch command line parsing from llvm::cl to OptTable (byproduct: drop -long-option?)

llvm/tools/ include some binary utilities used as replacement for GNU
binutils, e.g. llvm-objcopy, llvm-symbolizer, llvm-nm.
In some old threads people discussed some drawbacks of using cl::opt for
user-facing utilities (I cannot find them now).
Switching to OptTable is an appealing solution. I have prepared two patches
for two binary utilities: llvm-nm and llvm-strings.

* llvm-strings https://reviews.llvm.org/D104889
* llvm-nm https://reviews.llvm.org/D105330

llvm-symbolizer was switched last year. llvm-objdump was switched by thakis
earlier this year.

The switch can fix some corners with lib/Support/CommandLine.cpp. Here is a
summary:

* -t=d is removed (equal sign after a short option). Use -t d instead.
* --demangle=0 (=0 to disable a boolean option) is removed. Omit the option
or use --no-demangle instead.
* To support boolean options (e.g. --demangle --no-demangle), we don't need
to compare their positions (if (NoDemangle.getPosition() >
Demangle.getPosition()) , see llvm-nm.cpp)
* grouped short options can be specified with one line
`setGroupedShortOptions`, instead of adding cl::Grouping to every short
options.
* We don't need to add cl::cat to every option and call
`HideUnrelatedOptions` to hide unrelated options from --help. The issue
would happen with cl::opt tools if linker garbage collection is disabled or
libLLVM-13git.so is used. (See https://reviews.llvm.org/D104363)
* If we decide to support binary utility multiplexting (
https://reviews.llvm.org/D104686), we will not get conflicting options. An
option may have different meanings in different utilities (especially for
one-letter options).

*I expect that most users will not observe any difference.*

There is a related topic whether we should disallow the single-dash
`-long-option` form.
(Discussed in 2019:
https://lists.llvm.org/pipermail/llvm-dev/2019-April/131786.html Accept
--long-option but not -long-option for llvm binary utilities)
*I'd like to disallow -long-option but may want to do this in a separate
change.*
The main point is that (1) grouped short options have syntax conflict with
one-dash long options. (2) the GNU getopt_long style two-dash long option
is much more popular.

I can think of potential pushback for some Mach-O specific options, e.g. nm
-arch
http://www.manpagez.com/man/1/nm/osx-10.12.6.php says `-arch` has one dash.
If such options may have problems, we can keep supporting one dash forms.
With OptTable, allowing one-dash forms for a specific option is easy.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20210702/5f2fe78f/attachment.html>

On Fri, Jul 2, 2021 at 10:15 AM Fāng-ruì Sòng via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> llvm/tools/ include some binary utilities used as replacement for GNU
> binutils, e.g. llvm-objcopy, llvm-symbolizer, llvm-nm.
> In some old threads people discussed some drawbacks of using cl::opt for
> user-facing utilities (I cannot find them now).
>
Would be good to describe some of the known drawbacks/expected benefits.

One potential one (though I don't recall it being discussed recently) would
be that maybe this addresses the issue of global ctors in cl::opt? Does
OptTable avoid/not use global constructors? That would be nice - it's an
ongoing issue that LLVM library users pay for command line argument support
they have no need for in the form of global ctor execution time.

> Switching to OptTable is an appealing solution. I have prepared two
> patches for two binary utilities: llvm-nm and llvm-strings.
>
> * llvm-strings https://reviews.llvm.org/D104889
> * llvm-nm https://reviews.llvm.org/D105330
>
> llvm-symbolizer was switched last year. llvm-objdump was switched by
> thakis earlier this year.
>
> The switch can fix some corners with lib/Support/CommandLine.cpp. Here is
> a summary:
>
> * -t=d is removed (equal sign after a short option). Use -t d instead.
> * --demangle=0 (=0 to disable a boolean option) is removed. Omit the
> option or use --no-demangle instead.
> * To support boolean options (e.g. --demangle --no-demangle), we don't
> need to compare their positions (if (NoDemangle.getPosition() >
> Demangle.getPosition()) , see llvm-nm.cpp)
> * grouped short options can be specified with one line
> `setGroupedShortOptions`, instead of adding cl::Grouping to every short
> options.
> * We don't need to add cl::cat to every option and call
> `HideUnrelatedOptions` to hide unrelated options from --help. The issue
> would happen with cl::opt tools if linker garbage collection is disabled or
> libLLVM-13git.so is used. (See https://reviews.llvm.org/D104363)
> * If we decide to support binary utility multiplexting (
> https://reviews.llvm.org/D104686), we will not get conflicting options.
> An option may have different meanings in different utilities (especially
> for one-letter options).
>
> *I expect that most users will not observe any difference.*
>
> There is a related topic whether we should disallow the single-dash
> `-long-option` form.
>
(Discussed in 2019:
> https://lists.llvm.org/pipermail/llvm-dev/2019-April/131786.html Accept
> --long-option but not -long-option for llvm binary utilities)
> *I'd like to disallow -long-option but may want to do this in a
separate
> change.*
>
I'd say definitely do this as a separate change. I expect there'd be a
long
tail of users after this change ships in an LLVM release, etc, such that we
may want to undo some amount of it a long time after the change is made.

>
> The main point is that (1) grouped short options have syntax conflict with
> one-dash long options. (2) the GNU getopt_long style two-dash long option
> is much more popular.
>
> I can think of potential pushback for some Mach-O specific options, e.g.
> nm -arch
> http://www.manpagez.com/man/1/nm/osx-10.12.6.php says `-arch` has one
> dash.
> If such options may have problems, we can keep supporting one dash forms.
> With OptTable, allowing one-dash forms for a specific option is easy.
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20210702/9b784f4f/attachment.html>

On 7/2/21 10:14 AM, Fāng-ruì Sòng via llvm-dev wrote:
> llvm/tools/ include some binary utilities used as replacement for GNU 
> binutils, e.g. llvm-objcopy, llvm-symbolizer, llvm-nm.
> In some old threads people discussed some drawbacks of using cl::opt 
> for user-facing utilities (I cannot find them now).
> Switching to OptTable is an appealing solution. I have prepared two 
> patches for two binary utilities: llvm-nm and llvm-strings.
>
> * llvm-strings https://reviews.llvm.org/D104889 
> <https://reviews.llvm.org/D104889>
> * llvm-nm https://reviews.llvm.org/D105330 
> <https://reviews.llvm.org/D105330>
>
> llvm-symbolizer was switched last year. llvm-objdump was switched by 
> thakis earlier this year.
>
> The switch can fix some corners with lib/Support/CommandLine.cpp. Here 
> is a summary:
>
> * -t=d is removed (equal sign after a short option). Use -t d instead.
> * --demangle=0 (=0 to disable a boolean option) is removed. Omit the 
> option or use --no-demangle instead.
To me, removing these would make the interface *worse*.  This is purely 
subjective, but I use the second item regularly when locally debugging 
to swap back and forth between two modes easily.
> * To support boolean options (e.g. --demangle --no-demangle), we don't 
> need to compare their positions (if (NoDemangle.getPosition() > 
> Demangle.getPosition()) , see llvm-nm.cpp)
> * grouped short options can be specified with one line 
> `setGroupedShortOptions`, instead of adding cl::Grouping to every 
> short options.
> * We don't need to add cl::cat to every option and call 
> `HideUnrelatedOptions` to hide unrelated options from --help. The 
> issue would happen with cl::opt tools if linker garbage collection is 
> disabled or libLLVM-13git.so is used. (See 
> https://reviews.llvm.org/D104363 <https://reviews.llvm.org/D104363>)
> * If we decide to support binary utility multiplexting 
> (https://reviews.llvm.org/D104686
<https://reviews.llvm.org/D104686>),
> we will not get conflicting options. An option may have different 
> meanings in different utilities (especially for one-letter options).
>
> *I expect that most users will not observe any difference.*
>
> There is a related topic whether we should disallow the single-dash 
> `-long-option` form.
> (Discussed in 2019: 
> https://lists.llvm.org/pipermail/llvm-dev/2019-April/131786.html 
> <https://lists.llvm.org/pipermail/llvm-dev/2019-April/131786.html> 
> Accept --long-option but not -long-option for llvm binary utilities)
> *I'd like to disallow -long-option but may want to do this in a 
> separate change.*
> The main point is that (1) grouped short options have syntax conflict 
> with one-dash long options. (2) the GNU getopt_long style two-dash 
> long option is much more popular.
>
> I can think of potential pushback for some Mach-O specific options, 
> e.g. nm -arch
> http://www.manpagez.com/man/1/nm/osx-10.12.6.php 
> <http://www.manpagez.com/man/1/nm/osx-10.12.6.php> says `-arch` has 
> one dash.
> If such options may have problems, we can keep supporting one dash forms.
> With OptTable, allowing one-dash forms for a specific option is easy.
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20210705/47122f29/attachment.html>