Enhancement: simplify mkiocccentry command line

[xexyl]

Is there an existing issue for this?

• I have searched for existing issues and did not find anything like this

Describe the enhancement

In a number of places it has been suggested that for POST IOCCC28 the mkiocccentry tool could have a command line syntax change.

One of the comments is here but there were also some in the issue #1060 (which mostly seems complete now that we decided we will have this part in a new issue post IOCCC28).

The idea is that instead of having to specify prog.c, Makefile, remarks.md and then any extra data files, one can just specify a directory of files that will be copied to the work directory (by mkiocccentry), except dot files of course.

Relevant images, screenshots or other files

n/a

Relevant links

n/a

Anything else?

--

TODO before this can be closed

• verify updates to FAQ, guidelines and rules to account for the chkentry arg order being changed are okay
• Address comment 2592821650

See also GH-issuecomment-2622887502.

For file permissions set GH-issuecomment-2625020203.

[SirWumpus]

I think the idea of specifying a "staging" directory is useful. Lots of packaging tools for OS binaries and similar use the concept of "staging". This might be easier for contestants to manage.

[xexyl]

I think the idea of specifying a "staging" directory is useful. Lots of packaging tools for OS binaries and similar use the concept of "staging". This might be easier for contestants to manage.

Okay in that case I guess the option could mean to NOT do it this way.

I can see how it might be useful this way. Although I am not sure atm how we will do it skipping dot files unless we use a DIR * but I will think about that later.

And maybe if we have time before the contest opens I can implement this feature before.

Do you think that my proposed option letter still works (keeping in mind that many letters are already used)?

..

As an aside: one could just use in the command line as it stands the usual required files and then foo/*.

Honestly I am not really sure why I don't know how much I like this idea as my development Makefiles have a submit rule which copies the necessary files to its own subdirectory and thus would prevent the problem.

Perhaps it's because it makes an assumption that everyone has same system.

[xexyl]

As for implementation details.

It would be nice if the tool could do:

cp -r dir/*

but the problem is when it reaches a subdirectory if that subdirectory has a dot file it would copy that over too.

So a question has to be answered.

Since this new idea would assume that all files in the directories should be copied over we could simply have the relevant tools flag any invalid dot files. txzchk does this for example.

This would greatly simplify programming it something like above though it would still have to search for the required files.

But if we wanted to be 'kinder' we could skip dot files.

My suggestion is that we don't skip them and this is why.

Right now the user can specify dot files too. They will be flagged.

However if we copy the directory over recursively we will have to either use something like DIR * or else let the other tools flag the error.

I believe that we put in chkentry this code as well but right now mkiocccentry does the check on all files.

It might be nice to have it still do that. But on the other hand we might say that the chkentry (and also txzchk) is the appropriate tool: mkiocccentry gathers the files and also does some checks like on the required files but it could be argued the rest are better checked by the other tools.

Since the tarball is not formed before chkentry validates it it might not be a problem.

The only benefit of having mkiocccentry do it is that it does the checks BEFORE asking the user for input. It would be super annoying to miss a dot file (or even not knowing they're not allowed) and then inputting all that information only to have to redo it.

So on second thoughts I guess the tool should also validate the files.

I guess what this means is it should use DIR *.

What do you both think? Am I not thinking of a simpler way to go about it? I do think it's better we check. However I think it might be better to not just skip invalid files but rather make it an error. This way they are aware of it: what if they expect a certain file to be there but then it's not packaged? That would make them lose!

Have to go afk again. Back later!

[lcn2]

The tool will need to use the nftw(7) facility to walk the topdir directory and evaluate the file type, filename, path, path depth: building up a dynamic array list of things found and counting along the way to be sure there are not too many files.

It should NOT call the shell and use the cp(1) command!

The tree walking code will need to remember the Makefile, prog.c, and remarks.md files, and do the appropriate checks when these required files are found.

UPDATE 0

Removed the quoted text.

[xexyl]

As for implementation details.
It would be nice if the tool could do:

cp -r dir/*

but the problem is when it reaches a subdirectory if that subdirectory has a dot file it would copy that over too.
So a question has to be answered.
Since this new idea would assume that all files in the directories should be copied over we could simply have the relevant tools flag any invalid dot files. txzchk does this for example.
This would greatly simplify programming it something like above though it would still have to search for the required files.
But if we wanted to be 'kinder' we could skip dot files.
My suggestion is that we don't skip them and this is why.
Right now the user can specify dot files too. They will be flagged.
However if we copy the directory over recursively we will have to either use something like DIR * or else let the other tools flag the error.
I believe that we put in chkentry this code as well but right now mkiocccentry does the check on all files.
It might be nice to have it still do that. But on the other hand we might say that the chkentry (and also txzchk) is the appropriate tool: mkiocccentry gathers the files and also does some checks like on the required files but it could be argued the rest are better checked by the other tools.
Since the tarball is not formed before chkentry validates it it might not be a problem.
The only benefit of having mkiocccentry do it is that it does the checks BEFORE asking the user for input. It would be super annoying to miss a dot file (or even not knowing they're not allowed) and then inputting all that information only to have to redo it.
So on second thoughts I guess the tool should also validate the files.
I guess what this means is it should use DIR *.
What do you both think? Am I not thinking of a simpler way to go about it? I do think it's better we check. However I think it might be better to not just skip invalid files but rather make it an error. This way they are aware of it: what if they expect a certain file to be there but then it's not packaged? That would make them lose!
Have to go afk again. Back later!

The tool will need to use the nftw(7) facility to walk the topdir directory and process each file in turn.

Okay that's an interesting idea. I'll have to figure out the callback semantics then. Happily there's an example in the man page in linux. Fortunately this will also be after the next IOCCC. Unfortunately it seems it's going to be necessary anyway for the subdirectories, though the way I brought up over there is probably simpler.

Anyway I can't deal with that today. I will look at it tomorrow.

[xexyl]

.. so with that in mind please do the feature request issue if you can do it today.

[xexyl]

.. actually looking at the example and in particular the struct it seems more complicated. But I only briefly looked. I'll have to look more and check into it.

I might add it seems to require a feature test macro in case you did not know.

But you added top priority not post IOCCC28? Does that mean you'd like it before the next one?

[lcn2]

We are not sure about the -r as this just adds to the complexity of the command line. Only supporting:

mkiocccentry workdir topdir

Seems much more straightforward.

[xexyl]

We are not sure about the -r as this just adds to the complexity of the command line. Only supporting:

mkiocccentry workdir topdir

Seems much more straightforward.

I already gave reasons this is not necessarily a good idea. It makes it a big assumption that they want all the files in that directory to be added.

For instance in my submissions I have many more files than what I submit. Sure I make a Makefile rule that copies the right files into the submit directory but what if I forget to copy something over to that directory? Without typing it explicitly I might miss that something is missing.

I realise that more people might like it the other way which is why I think it should be the default but there should be a way to disable that for those who want more control.

[SirWumpus]

We are not sure about the -r as this just adds to the complexity of the command line. Only supporting:

mkiocccentry workdir topdir

Seems much more straightforward.

I already gave reasons this is not necessarily a good idea. It makes it a big assumption that they want all the files in that directory to be added.

Which is why one would probably create a "staging" directory to drop all the goodness required for submission into one place to make sure no sundries appear.

Update 0

The Contestant has to be responsible for some organisation on their part IMO.