Hi Paul,
Well….
> On Sep 6, 2022, at 12:18 AM, Paul Dufresne via Freedos-devel
> wrote:
>
> I wrongly said:
> >It's probably ok... because fdnpkg probably won't too.
> Nonsense! fdnpkg have no way to know that many directories have been combined
> into one.
>
> So I just need fdpkg to accept to combine directories with same 8 chars
> truncated length to be combined together.
If the DJGPP program (DJ_GPP) works fine when the long directory names are
truncated and combined, a possible solution would be to create a utility (or
add an option/modify fdpkg.sh) to perform the task. Basically, scan directory
names and if the would be truncated to 8.3 chars (and collide with another
directory), to just do that in advance.
I don’t know how familiar you are with bash scripting. However, the fdpkg.sh
shell script was something thrown together quickly and is not very complex.
A general overview of what happens when it is executed…
First, it sets a couple internal global variables. Everything else is broken
into functions to make it more organized. Then it calls the PKGU_main function
with the command-line passed to the script.
PKGU_main - parses the command line options and either expands zips or tries to
compress paths into archives. When compressing and it is LFN mode is enabled
(the default, set by "LFN_ZIP=true”), it will call the a function
(PKGU_SmashLFN) to see if any LFNS are present and any LFNFILES.ZIP archives
are required. Otherwise, it will just proceed without checking. Next, it will
attempt to compress the path as DOS 8.3 compatible. If this cannot be done, the
script will abort with an error.
PKGU_SmashLFN - looks at the the primary directory names and determines if they
are SOURCE, NORMAL or SPECIAL directories (like LINKS). Those are either
handled like sources (PKGU_CheckSources), the binaries (PKGU_CheckPath) or
specifically exempt from LFN scanning and compression.
PKGU_CheckSources and PKGU_CheckPath - Basically do the same things in a
slightly different way with with slightly different settings. The search for
the LFNs and Compress those when found. Their are slightly different conditions
used by each to determine what qualifies as a Long File Name. PKGU_CheckPath
uses the PKGU_SearchLFN to scan for any files that are NOT compatible with 8.3.
However, PKGU_CheckSources uses PKGU_SearchEFN that will do the same but
include any filename that is not uppercase.
When the RBE performs the compression of packages, it does some similar
scan/compressing. However for several reasons, it always compresses the SOURCE
directory. Primarily, this tends to reduce the overall size of the archive. It
also requires less disk space when sources are installed. A side benefit is
that when always compressing that path, it never needs to actually scanned. I
should update fdpkg.sh to just always compress those files without scanning to
make it more like process used by the RBE.
So….
There are several ways to add a TRUNCATE and COMBINE long path names to the
script. First, probably add a global variable to en/disable allowing the
action. Something like LFN_CRUSH=false. Then a command line option like ‘-c’ to
enable and maybe ‘-n’ to disable it. (default probably -c)
Then under PKGU_SmashLFN for specific paths call a new function to perform the
action. Or, do that under PKGU_CheckSources and/or PKGU_CheckPath. The new
could be called PKGU_CrushDirs.
PKGU_CrushDirs could be implemented several ways. The easiest would be to just
figure out what the truncated directory name would be and see if that collides
with anything else. When that happens, just move that stuff into the same
truncated directory.
If that sounds like it would solve the problem, I could add it to the fdpkg.sh
script. Then if it works, implement the equivalent in the RBE.
:-)
Jerome
>
> ___
> Freedos-devel mailing list
> Freedos-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/freedos-devel
___
Freedos-devel mailing list
Freedos-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/freedos-devel
