1
2 09/21/87 hierarchy_backup
3 Known errors in the current release of hierarchy_backup.
4 # Associated TR's
5 Description
6
7 49 phx20752
8 backup_dump verified
9 timing problem allowing backup_dump to abort when it is called
10 via an alarm wakeup and a typed ahead call.
11
12 48 phx20313
13 backup_map_ verified
14 bakup_map_$error_line is called with too many arguments.
15
16 46 phx20013
17 reload investigating >udd>TR>TR_tc>tc1>explanatory.trans.phx20013
18 reload does not trim directories that are not on tape, but on
19 disk and have not been modified since the dump was
20 taken. .
21
22 As time and resources permit this problem will be evaluated and better
23 defined at a future date.
24
25 43 phx20082
26 The hierarchy reloader is unable to reload absentee queues from ring 1
27 level.
28
29 This problem will be investigated after MR12 is released.
30
31 42 phx17931
32 The hierarchy reloader may be setting acls incorrectly when reloading.
33
34 once investigated, more information will be added here.
35
36 40 phx19804
37 The TR describes a symtom of the actual problem. The problem being
38 that the hierarchy dumper creates all its work space as static
39 variables and does not initialize any of them. The programs need to be
40 changed to set all variables to a proper initial value.
41
42 39 phx15671
43 the hierarchy dumper programs contain declarations for system
44 subroutines which are obsolete. These declarations need to be updated
45 to match the parameter declarations of the current routines.
46
47 34 phx14430
48 Retriever/reloader spuriously sets ring brackets to 1 on containing dir
49 of a reloaded directory. That is, if you reload >a>b>c, and >b does
50 not exist, it is created with ring brackets of 1,1.
51
52 31 phx18476
53 backup_load does not clean up after itself. In particular, on tape
54 errors, it is very good at not detaching its switches.
55
56 29 phx17583
57 The complete_dump command fumbles the dump suffix within the -control
58 control arg.
59
60 25 phx15734
61 The hierarchy dumper requires "s" access to the directory containing
62 the dump control file. This should be unneccessary.
63
64 22 phx15616
65 The retriever should do something to the quota on master directories it
66 loads into when cross-retrieving non-mdirs into it.
67
68 21 phx15802
69 A record quota overflow occuring during the backup_dump preattach leads
70 a loop through various routines.
71
72 19 phx15859
73 The reloader and dumper should use the user's name in the dprint header
74 args, instead of making everyone's output labeled "the same".
75
76 18 phx15348
77 On rare occasions, the dump maps from the reloader -noreload will
78 miss a couple segments that are on the tape and reloadable.
79
80 17 phx14714
81 Cross-retrieving under a new name attempts to add all the old names to
82 the new object, deleting an old copy of it under the old name.
83
84 That is, if the old name is FOO, and the new name FOO.NEW, it tries to
85 add FOO to FOO.NEW, defeating the purpose of the cross-retrieve.
86
87 16 phx14707
88 you cannot turn off the map for incremental dumps. The assignment of
89 "1"b to the mapsw in start_dump.pl1 is in the wrong place.
90
91 15 phx14544
92 The dumper and reloader conspire to screw up the top level dir of a
93 retrieved hierarchy. If >udd>a is an upgraded directory, and it is
94 dumped and then deleted, and then a retrieval is made for >udd>a>b, a's
95 attributes will not be correctly set when it is created if it was the
96 top level directory of the dump. In the upgraded case this causes
97 quota inconsistencies and SOOS directories, and prevents the
98 subdirectory from being reloaded.
99
100 13 phx13832
101 Crossretrievals of level 1 directories do not work.
102
103 12 phx13773
104 The dumper ignores the volume dumper switches. Also the DNZP switch. Possibly other things. The fix requires a rewrite of the dumper, with attendent compatability difficulties.
105
106 10 phx11581
107 reload -nomap -noreaload produces a map anyway.
108
109 7 phx13510
110 Complete_dump -nomap -other things still produces a map.