PR#340 alternative ? flush -noconfirm .. exit exit_status -noconfirm

[dlmiles]

@wulffern please advise, test as alternative to PR#340

Documented hidden/secret exit [-noconfirm] option to disable exit prompt Added exit [exit_status] option to send exit_status number Added flush -noconfirm option to disable flush prompt

I have tested the 'exit' patches.

I have not yet tested the 'flush' patch. Still comment back when testing complete on my side but would appreciate feedback if it closes your concern.

[wulffern]

@dlmiles Yup, I can confirm that the flush -noprompt does what I want. Thanks.

It's a bit more finicky to use since I need to know that the -noprompt option is in the Magic version.

Magic throws an error if I use -noprompt in a Magic version that does not have it, but with a "flush_all" script like the below it works fine.

set values [ cellname list all ]
set mrev [lindex [lreverse [split [exec magic --version] "."]] 0 ]
foreach x $values {
    if {${mrev} > 495} {
        flush $x -noprompt
    } else {
        flush $x
    }
}

PS: I could not find the revision in any global tcl variable, nor was I able to redirect "version" to a variable, but I'm sure there is a more elegant way to do the above.

[RTimothyEdwards]

@wulffern : catch {flush -noprompt} should work around the backwards-compatibility failure, or maybe

if {[catch {flush -noprompt}]} {flush}

[dlmiles]

@RTimothyEdwards Did you want a revision using Lookup() ? not clear as it PR marked approved.

@wulffern maybe you can setup $ENV{'MAGIC_VERSION'} before magic is run, to TCL inherits envars, like in a startup wrapper shell script ? This way it is always available. This isn't much difference to how you are doing it now really. Maybe there should be a patch to make such a thing available from TCL an auto-populated variable.

Maybe it can be run as flush $x -noprompt and if this errors, too many args, it just retries the command without -noprompt. This may not be a possible if $x is not present, as -noprompt' maybe processed as a cellname` which will probably also error.

Difficult to provide a backward compatible mechanism that isn't as finicky, as least you can make it into myflush TCL proc to clean / hide this extra work?

Your original patch with an environment variable set from inside (or maybe inherited from outside) might be better reworked with the batch_mode feature, such that you can start up in GUI mode with batch_mode and it has the effect of making all interactive decision dialogs to cause error termination. It also make sense for $ENV{'MAGIC_BATCH_MODE'} to also be a thing so scripts a runtime. Maybe this is already possible.

Which then necessitates an option like -noprompt for every command that might present a dialog, which allows the command to succeed. This is my use case, I am looking to understand how some test cases maybe prepared that operate in GUI mode, but I need a way to manage dialog interactions and terminate with exit_status. Hence this PR.

[RTimothyEdwards]

@dlmiles : I approved it because the code is usable as-is, but if you can change it to use Lookup() I'd appreciate it.

FYI: You can use the version command in magic to get the version; it does not need to come from an external variable. Although a version --list option would be nice to get just the version numbers back as a Tcl list instead of text.

[dlmiles]

I have checked over the Lookup() methods and the usages in the codebase I can't quite see an obvious way all the requirements can be met :

• allow -option to be any order (ok regular Lookup usage does this part)
• continue to ensure -nodef is the same as the longer form -nodeference but also -nodefer etc.. to ensure a users script doesn't break (this is the main blocker)
• extract the [exit_status] option, is this allowed to be in any position ? (unclear how the remaining arguments left in the list, unclaimed by any Lookup is extracted into a canonicalized form argv[]) so the arg count in this category can be seen, to allow exit -noprompt 42 but also check for invalid exit -noprompt 42 0
• error on weird usage like exit 1 -noprompt -noprompt -nodef -nodeference maybe this is a minor one

Re current patch having fixed order, sure I added in a way that should not break any current users scripts and documentation indicates correct order to make it as easy to discover and get working as possible.

[RTimothyEdwards]

Pulled and merged into the opencircuitdesign.com code base.

[dlmiles]

Not sure what happened to this change set it did not make it into master

Instead there is 772bfe2f71f31bf8e0254fe685c8bc65e0829709 from Carsten (and the initial version that was using a non-standard method to activate the feature)

So I rebase 772bfe2f with a revert ahead of the 3 commits here.

[dlmiles]

@RTimothyEdwards No re-open issue button here.

Just in case it is not clear from issue information which 4 commits they can be seen here https://github.com/RTimothyEdwards/magic/compare/master...dlmiles:magic:master-upstream-20241009-noconfirm the new HEAD is 0a27045

Update: This is also the cause of the WASM build failures in GHA. the commit from #340 exposes magicinterp without ifdef for WASM use. So the 4 commits now here will resolve that as well.