On 2018-03-05 16:39, Andreas Reichel wrote: > On Mon, Mar 05, 2018 at 02:40:40PM +0100, Jan Kiszka wrote: >> On 2018-03-05 14:08, Andreas Reichel wrote: >>> On Mon, Mar 05, 2018 at 01:43:01PM +0100, [ext] Jan Kiszka wrote: >>>> On 2018-03-05 13:24, Andreas Reichel wrote: >>>>> On Mon, Mar 05, 2018 at 01:01:14PM +0100, Jan Kiszka wrote: >>>>>> Hi Andreas, >>>>>> >>>>>> what is bg_setenv without any parameters supposed to do, and what does >>>>>> it actually - besides printing a suspicious "Environment update was >>>>>> successful."? >>>>>> >>>>> >>>>> Good point, thanks for asking. >>>>> >>>>> If you specify no parameter, the program goes into >>>>> (write_mode), since you call bg_setenv and not bg_printenv. >>>>> >>>>> Output to file is false, since no "-f" given. >>>>> >>>>> auto_update is false, since no "-u" given. >>>>> >>>>> part_specified is false, since no "-p" given. >>>>> >>>>> This means, the program opens the latest environment and then calls >>>>> update_environment(env_new), where env_new is a handle to the latest >>>>> environment. (nothing new created at this point). >>>>> >>>>> update_environment then processes an empty task list which does not >>>>> change the latest environment. It then recalculates the CRC32 again. >>>>> >>>>> The program then stores the unmodified environment together with the >>>>> recalculated CRC32 back into the same place and exits. >>>>> >>>>> Thus, bg_setenv without parameters performs a NOP on the current >>>>> environment data except the CRC and writes the values back again. >>>>> >>>>> Thus, the question here is "what is the default behavior, a program >>>>> should have, if started without any parameter". >>>> >>>> Right, and what is the use case behind it. >>>> >>>>> >>>>> I would say, in case of bg_setenv, the behavior is not offbeat, the >>>>> message is true and if the user has doubts about what to do, he should >>>>> use --help anyway. Per definition, all arguments are optional, therefore >>>>> it would be strange if the help page was printed as if the user missed >>>>> something, since optional arguments are not required per definition. >>>> >>>> Normally, one model such a tool via subcommands: >>>> >>>> tool subcommand [OPTIONS] >>>> >>>> Unless there is a case for the current default behaviour. But it seems >>>> like it is rather a side effect of the normal use cases. >>>> >>> >>> In the very beginning the idea was to have the 'bg_setenv/printenv' pair >>> modelled like 'fw_setenv/printenv' from U-Boot, where no subcommands >>> are used. >>> >>> If we use sub-commands anyway, it does not make sense to have two >>> different names for the program but rather a >>> bg_envtool or so with >>> >>> bg_envtool set >> >> It rather seems we would then need bg_envtool set <subsubcommand> - I >> don't think that is useful. >> >> We currently have bg_setenv -c | -i | -s | -u, right? That could become >> bg_setenv confirm_update | start_update | ... and keep bg_printenv. At >> least we should define which switches are at least needed in order to do >> something useful (still considering the parameter-free invocation useless). > > Well... start_update does not make sense from the tool's point of view.
If the tool is too low-level for such an abstraction, well. It's usage is definitely not intuitive yet to guide the user through typical workflows. > > There already are long names for the switches. And one of the > advantages is that they can be combined... which is not the case if > we change them into sub-commands. You can always abbreviate long subcommands after a sufficient amount of distinguishing characters. Common pattern in many tools. And/or provide bash_completion. > > For example, one can use -u -x test=1 -c to automatically create a new > environment revision, set a user variable and immediately confirm > without further testing. > > Or if we want to simulate a failed update finalization on config > partition number 0: > -p 0 -k kernel-name -c -i 1 > > which sets in_progress after confirming to see how the bootloader > reacts. > > I agree in any case to solve this strange behavior (calling without > any arguments) by outputting "Nothing to do. Please specify at least > one optional parameter (see --help for further information)." That would already improve the situation, indeed. > > But, categorizing into sub-commands does > * limit our usage possibilities That is surely not true when the interface is designed to accept enough tweaking parameters in the various sub-commands. > * increase code complexity > * break the way the tool is used * simplify its usage Jan -- You received this message because you are subscribed to the Google Groups "EFI Boot Guard" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/efibootguard-dev/bed04951-bf5d-fef3-94d2-d22947b6dc1f%40siemens.com. For more options, visit https://groups.google.com/d/optout.
