vsiftp.vmssoftware.com/[e-mail]/FREEWARE/FREEWARE40.ZIP/flistfrontend/src/flist.bug

Text File | 1995-10-27 | 16.0 KB | 244 lines

I have a hook for 'lib$get_symbol' in 'dirarg', which now tests for equates
into FLIST's own commands.  FLIST should (if I figure how to do error-message
intercept) allow any command to be spawned which doesn't fall into its own
tables.  (Note: It should be able to execute a DCL-procedure as the current
entry by typing "@/ ", followed by appropriate parms.)

The spawn-command supports a crude suspended-process operation.  Should I
also disable control-Y?  (Must remember to re-init terminal, though)

RMS notes: look into UBI (shared access), RRL (read record locked), and
block I/O (only if I did my own COPY).

Use of 'DIRPROT' is logically correct, but too slow if a wildcard is specified.
Should either simply verify existence (keep your fingers crossed), or put
a "working..." sign out.

The completion routine stuff I added to DDS.C is ok, but I must disable
the messages still.  How do I interrogate/clear/set these flags (SET MESSAGE).

In FLCOPY, there is a problem with copying to lower version than an
existing file: COPY emits an error message unless I put an explicit version
(e.g., the version I would get anyway).

I fixed some of the spacing needs of DCLARG ("/ opt = val"), but I note
that this is primarily to fill in the needs for symbol substitution.
DCL itself allows a space whenever a punctuation mark is pending.

(8-Aug-84) I ran FLIST on "SYS$EXAMPLES:", which SYS$PARSE expanded to
"SYS$SYSROOT:[SYSHLP.EXAMPLES]".  Then I did a SPAWN into that directory.
The SHOW-DEFAULT command showed the expanded string, but I was unable
to do "DI" (DIR was ok), or "FORT", until I re-issued "SET DEF SYS$EXAMPLES:".
A subsequent SHOW-DEFAULT showed no difference (???).

(17-aug-84) Implemented wildcard delete/purge.  For consistency, should also
permit copy(?).

(15-aug-84) In cleanup, I should be more scrupulous about checking the
result from calloc/realloc.  I might get burned sometime.  Also, I should
get rid of constants for buffer sizes (80, 132, 256), replacing them with
MAX_CRT and MAX_PATH/NAM$C_MAXRSS) as appropriate.

03-sep-84: Should get rid of procedure-call to TOUPPER, TOLOWER to reduce
code some more.

06-sep-84: Not sure what to do about '"' in DCLARG, since it is stripped
before unix-argv interface gives it to me.

23-oct-84: The PURGE command can be confused by multiple directories,
apparently inheriting the pathname from the current default, rather than
properly testing for explicit-pathname (see: DIRFIND.C).

14-nov-84: If a file is in a directory, but does not exist, then DIRECTORY
will give the message "no such file".  This corresponds to the code
RMS$_FNF, which FLIST uses as a magic code for deleted-file.  Should use
instead a code which isn't used by RMS.

12-dec-84: If I used my own parse (instead of SYS$PARSE) in DIRENT_CHOP,
I could cut directory-read by 20% when going across decnet.  However,
DIRECTORY itself is about 5 times faster than FLIST. ???
(For this, test NAM.NAM$NOP |= NAM$M_SYNCHK).

13-dec-84: FLIST is robust enough that I should put in ^C, ^Y intercepts.

19-dec-84: It would be nice to have a control which can abort reading a long
directory without killing the entire program.

21-dec-84: It would be nice (in DELETE and PURGE) to update the number-of-files
field in the status line, so that the user can see that something is happening.
FILENT is currently 60 bytes.  If I added another 4 bytes this would
make page boundaries even.  (E.g., record-size)

25-dec-84: The current major problems are:

    - Find how to make this run faster on DECNET.
    - Find how to fix the bug in VMS quota, and to display this in my
      program.

29-dec-84: Should permit a filespec argument for "?SIZE".
In general, should get rid of unnecessary parse/search.

30-dec-84: 'crt_refresh' assumes that the cursor position is not near home.

08-jan-85: '/NOVERSIONS' changes only the default on directory read.
This is a hole.  Should test the fnb on search to ensure that we don't
put spurious versions into display list.

10-jan-85: Would like to support VMS options:
    /EXCLUDE, /EXPIRED, /FILEID, /SELECT

Also, would like to permit BROWSE to run in (given VT100), or start-in (VT52)
a half-screen from FLIST.  This would make browsing much faster, since then
I could integrate it with a mode which ensures that the current-entry is
in the top half of the screen (problem if file is in the last ten entries!).

11-jan-1985: fixed scrolling-margin reset.  however, if user alters SET TERM
in a spawn, this affects the parent process also.

I would like to make the cursor invisible during screen refresh.  Can I
do this in ANSI?

28-jan-85: Running across decnet, the first SEARCH and the second and
following OPEN RMS-calls are the time-consuming ones.  It appears that
the first SEARCH-call caches some data for the OPEN.

04-feb-85: Should investigate lib-set-buffer for throughput.

05-feb-85: low-end aaa terminals do not support scrolling margins (i.e.,
non-ANSI).  There appears to be no way VMS can tell my programs about this.

08-feb-85: Is there any way to detect the current amount of stack usage
relative to the maximum?  If not, I must put a hard-coded limit in the
main program to limit recursion.  I would like to make the program tunable
in terms of the linked stack-limit.

In 'dirpath_rename', I am not checking for the use of the root-node
'[0,0]'.  The code is not conformable with 'pathdown'; probably need a
test for root-node.

23-mar-85:
Added display of file-id.  I noticed that at the root level, there are some
directories which are entered twice in the directory, once as a number, and
also as a name.  This is a hole in FLIST which I don't think I should try to
plug (yet).

26-mar-85:
Limited file-id to 5+1+5+1+1 = 13 columns.  Is the high-order word ever
nonzero?

How do I clear an unused AST?

29-mar-85:
Both FLIST and BROWSE should return the worst-error code, as well as printing
the error message.  (For example, FLIST does not now return an error code
if the first-level could not be read.)

Should use lib$get_ef instead of having hard-coded event flag numbers (see
'getraw') for qio's.

The out-band qio stuff will let me get at CTRL/C.  However, it does not
intercept while I am in help-library, and also it disables CTRL/C for
a subprocess.

13-apr-85:
/LOG should cover results of COPY and APPEND

26-apr-85:
If I get a message from system, etc., while in BROWSE, it gets caught on
the bottom of my screen as though it too were a scrolling region.
Maybe I should leave scrolling margins normally set up to bracket the
cursor-line, and enable them for scrolling only occasionally.

9-may-85:
Urbanski has a long-name file in his directory which cannot be found by
acp (because it was generated on v4.1).  Shouldn't stop directory-lookup,
but it does, because sys$search returns a no-more-files code immediately
after.  Why does DIRECTORY give a full list?

12-may-85:
'dirent' should give back only the worst error encountered, rather
than on each error (making entry to a new level nicer).

18-may-85:
Should extend /COLUMNS to permit user to strip off all columns (/fast).

30-may-85:
The (apparent) speed of FLIST could be much enhanced by doing SYS$OPEN
on demand (when we must first use the date, etc.) rather than on entry.
However, this would make the *total* slower, since we would have to do
SYS$SEARCH twice.  Should investigate this, to find how much time is
involved.  Only need to modify the flsort and dirent modules; can use a
convention of '0' in fstat for un-initialized data.

08-jun-85:
Should extend 'spawn' command to 'shell' (entry to DEC-shell).

10-jun-85:
Would like to show file-size in bytes; this is applicable to text files
(but not really to indexed files, which contain a toc).

Also, would be nice to have FLIST provide a mode (like the IBM program)
for selecting files from a list, return the list to a caller.
DCL doesn't have pipes/stack, must return this in variable(?)

Should provide a display command to show /CWIDTH (or put it in the
?COLUMNS display).

For the display-commands, would like to be able to use these as options.
To do this, consider making the option-interpretation of FLIST scan
left-to-right interpreting the commands in no-echo mode.

23-jun-85:
In going to CC2.0, encountered flakey-bug (inconsistent) in calls on 'sscanf'.
Coded around this with new module 'scanver'; ~10 hours of tests failed to
isolate *any* code errors.

Should put a latched-data into 'dirfind_tst'.  Now it is called only from
'dirread' and 'dirfind'.

26-jul-85:
Must suppress /DATE, etc., in options for edit-dir (subset).
Should provide entry for flcols to set column-list as an option-value,
i.e., either
    /c s d r
    /c=(s,d,r)
should be accepted by the scanner.

15-aug-85:
Should provide LOG, COM (CMD?) commands to permit dynamic use of FLIST.
Could maintain multiple log-files (put them on a linked list), one per level.

Should provide a mechanism for entering BROWSE from an FLIST command file
to permit it to revert to interactive mode for the duration of the
browse.

27-aug-85:
If we RENAME a directory, which in turn propagates into lengthened pathnames
for other entries, we do *not* increase 'ccolumns', since we do not scan the
'filelist[]' array to ensure that the lengthed pathnames are displayed.

16-sep-85:
STRWCMP does not check for the special case of '*' in a pathname conflicting
with the VMS convention (i.e., '[a*b]' should not match '[a.c.b]').

17-sep-85:
Looked across the net at ATCSDB::DRA0:[SYS0.], found that the netserver-log
file cannot be examined, since it is modified by observation.  However, some
versions will become "file-not-found" on SYS$OPEN, but are not counted
in deletions (probably because the dirent__read routine thinks that is a
bit too early).  Sometimes, also, I get an error from RMS (on UPDATE) saying
that the assign-channel service failed.  Could be either traffic-related,
or because I do not $CLOSE if $OPEN fails.

01-oct-85:
Must update help-file to show SET/SHOW substitution, as well as the order-of-
evaluation of synonyms.  Should permit FLIST's help command to accept argument.
Should modify EDTCMD module to treat <LF> same as in VMS 4.x

02-oct-85:
Should use qio-IO$_M_TYPEAHDCNT (p-39) in 'gotraw' rather than the qio with
timeout.

12-nov-85:
Should add optional arguments to SPAWN.  Also, add /ALL option to VERIFY.