[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: roland_libstore_modules_branch

From: Roland McGrath
Subject: Re: roland_libstore_modules_branch
Date: Mon, 11 Mar 2002 04:37:27 -0500 (EST)

> Ok, I bite.

Finally! ;-)

> First, it compiles and builds.

That much I knew.

> The statically compiled ext2fs contains the following store classes

You'll need device in there if you want to boot.  In my tree I have
../libstore/libstore_device.a linked into {ufs,ext2fs}.static, but I didn't
check those changes in anywhere since the branch is only in the libstore
directory.  I think we'll have a configure option for the set of store
types for the static filesystems.  

> The gunzip store is recognized, but I get an EIO error (I get this error
> with the old code, too, so I don't show it here).

Hmm.  We should look into that and make sure the gunzip and bunzip2 stores
work, since we want them for floppy booting.

> What does not work is a statically linked storeio.  It doesn't parse its
> argument!  argp calls the parsers with the ARGP_KEY_INIT (or whatever it was
> called), but does not parse any of the options on the command line.  Thus
> there is still some glitch.  I did test a storeio.static with the old code,
> and it worked.  So there is something funny going on.

You'll have to debug what is going on in argp.c then.  I can't really guess.

> I did only the above tests, not a full install and boot, which one should
> probably do next.  But it already looks very promising, fine work!

But it doesn't sound like you tested any of the dynamic loading
functionality at all.  It's true we don't need to test that before we can
put the code in, if it's not breaking anything that works now.  A simple
way to test is remove one of the modules like gunzip from the store-types
list and make libstore_gunzip.so.0.2, then test that -Tgunzip properly
loads the module.  To test the module loading for decode, you could do the
same with zero, installed as libstore_type-6.so.0.2 (6=STORAGE_ZERO) and
verify that storeread /dev/zero works to dynamically load the module.

But before those tests, if a full boot and tests on the system using the
new libstore all work, then I will merge the branch into the mainline.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]