Overview
| SHA1 Hash: | 8c6488ded2d656ba8e025c0abf9ed9bfaeab5007 |
|---|---|
| Date: | 2007-11-27 04:26:56 |
| User: | aku |
| Comment: | Continued work on the integrity checks for changesets. Moved callers out of transactions. Two checks are already tripping on bad changesets made by InitCSets (pass 5). |
| Timelines: | ancestors | descendants | both | trunk |
| Other Links: | files | ZIP archive | manifest |
Tags And Properties
- branch=trunk inherited from [a28c83647d]
- sym-trunk inherited from [a28c83647d]
Changes
[hide diffs]Modified tools/cvs2fossil/lib/c2f_integrity.tcl from [69cecd83d4] to [19b5dfccf5].
@@ -28,25 +28,31 @@
snit::type ::vc::fossil::import::cvs::integrity {
# # ## ### ##### ######## #############
## Public API
typemethod strict {} {
+ log write 4 integrity {Check database consistency}
+
set n 0
AllButMeta
Meta
return
}
typemethod metarelaxed {} {
+ log write 4 integrity {Check database consistency}
+
set n 0
AllButMeta
return
}
typemethod changesets {} {
+ log write 4 integrity {Check database consistency}
+
set n 0
- RevisionCSetLinkage
+ AllChangesets
RevisionChangesets
SymbolChangesets
return
}
@@ -54,11 +60,10 @@
## Internal methods
proc AllButMeta {} {
# This code performs a number of paranoid checks of the
# database, searching for inconsistent cross-references.
- log write 4 integrity {Check database consistency}
upvar 1 n n ; # Counter for the checks (we print an id before
# the main label).
# Find all revisions which disagree with their line of
@@ -269,11 +274,10 @@
}
proc Meta {} {
# This code performs a number of paranoid checks of the
# database, searching for inconsistent cross-references.
- log write 4 integrity {Check database consistency}
upvar 1 n n ; # Counter for the checks (we print an id before
# the main label).
# Find all revisions which disgree with their meta data about
@@ -289,26 +293,254 @@
;
}
return
}
- proc RevisionCSetLinkage {} {
+ proc AllChangesets {} {
+ # This code performs a number of paranoid checks of the
+ # database, searching for inconsistent changeset/revision
+ # information.
+
+ upvar 1 n n ; # Counter for the checks (we print an id before
+ # the main label).
+
+ # Find all revisions which are not used by at least one
+ # revision changeset.
+ Check \
+ {All revisions have to be used by least one revision changeset} \
+ {is not used by a revision changeset} {
+ -- Unused revisions = All revisions
+ -- - revisions used by revision changesets.
+ --
+ -- Both sets can be computed easily, and subtracted
+ -- from each other. Then we can get the associated
+ -- file (name) for display.
+
+ SELECT F.name, R.rev
+ FROM revision R, file F
+ WHERE R.rid IN (SELECT rid FROM revision -- All revisions
+ EXCEPT -- subtract
+ SELECT CR.rid FROM csrevision CR, changeset C -- revisions used
+ WHERE C.cid = CR.cid -- by any revision
+ AND C.type = 0) -- changeset
+ AND R.fid = F.fid -- get file of unused revision
+ }
+ # Find all revisions which are used by more than one revision
+ # changeset.
+ Check \
+ {All revisions have to be used by at most one revision changeset} \
+ {is used by multiple revision changesets} {
+ -- Principle of operation: Get all revision/changeset
+ -- pairs for all revision changesets, group by
+ -- revision to aggregate the changeset, counting
+ -- them. From the resulting revision/count table
+ -- select those with more than one user, and get their
+ -- associated file (name) for display.
+
+ SELECT F.name, R.rev
+ FROM revision R, file F,
+ (SELECT CR.rid AS rid, count(CR.cid) AS count
+ FROM csrevision CR, changeset C
+ WHERE C.type = 0
+ AND C.cid = CR.cid
+ GROUP BY CR.rid ) AS U
+ WHERE U.count > 1
+ AND R.rid = U.rid
+ AND R.fid = F.fid
+ }
+ # All revisions in all changesets have to agree on the LOD
+ # their changeset belongs to. In other words, all revisions in
+ # a changeset have to refer to the same line of development.
+ #
+ # Instead of looking at all pairs of revisions in all
+ # changesets we generate the distinct set of all LODs
+ # referenced by the revisions of a changeset, look for those
+ # with cardinality > 1, and get the identifying information
+ # for the changesets found thusly.
+ CheckCS \
+ {All revisions in a changeset have to belong to the same LOD} \
+ {: Its revisions disagree about the LOD they belong to} {
+ SELECT T.name, C.cid
+ FROM changeset C, cstype T
+ WHERE C.cid IN (SELECT U.cid
+ FROM (SELECT DISTINCT CR.cid AS cid, R.lod AS lod
+ FROM csrevision CR, revision R
+ WHERE CR.rid = R.rid) AS U
+ GROUP BY U.cid HAVING COUNT(U.lod) > 1)
+ AND T.tid = C.type
+ }
+ # All revisions in all changesets have to agree on the project
+ # their changeset belongs to. In other words, all revisions in
+ # a changeset have to refer to the same project.
+ #
+ # Instead of looking at all pairs of revisions in all
+ # changesets we generate the distinct set of all projects
+ # referenced by the revisions of a changeset, look for those
+ # with cardinality > 1, and get the identifying information
+ # for the changesets found thusly.
+ CheckCS \
+ {All revisions in a changeset have to belong to the same project} \
+ {: Its revisions disagree about the project they belong to} {
+ SELECT T.name, C.cid
+ FROM changeset C, cstype T
+ WHERE C.cid IN (SELECT U.cid
+ FROM (SELECT DISTINCT CR.cid AS cid, F.pid AS pid
+ FROM csrevision CR, revision R, file F
+ WHERE CR.rid = R.rid
+ AND F.fid = R.fid) AS U
+ GROUP BY U.cid HAVING COUNT(U.pid) > 1)
+ AND T.tid = C.type
+ }
+ # All revisions in a single changeset have to belong to
+ # different files. Conversely: No two revisions of a single
+ # file are allowed to be in the same changeset.
+ #
+ # Instead of looking at all pairs of revisions in all
+ # changesets we generate the distinct set of all files
+ # referenced by the revisions of a changeset, and look for
+ # those with cardinality < the cardinality of the set of
+ # revisions, and get the identifying information for the
+ # changesets found thusly.
+ CheckCS \
+ {All revisions in a changeset have to belong to different files} \
+ {: Its revisions share files} {
+ SELECT T.name, C.cid
+ FROM changeset C, cstype T
+ WHERE C.cid IN (SELECT VV.cid
+ FROM (SELECT U.cid as cid, COUNT (U.fid) AS fcount
+ FROM (SELECT DISTINCT CR.cid AS cid, R.fid AS fid
+ FROM csrevision CR, revision R
+ WHERE CR.rid = R.rid) AS U
+ GROUP BY U.cid) AS UU,
+ (SELECT V.cid AS cid, COUNT (V.rid) AS rcount
+ FROM csrevision V
+ GROUP BY V.cid) AS VV
+ WHERE VV.cid = UU.cid
+ AND UU.fcount < VV.rcount)
+ AND T.tid = C.type
+ }
+ return
}
proc RevisionChangesets {} {
+ # This code performs a number of paranoid checks of the
+ # database, searching for inconsistent changeset/revision
+ # information.
+
+ upvar 1 n n ; # Counter for the checks (we print an id before
+ # the main label).
+
+ # All revisions used by revision changesets have to refer to
+ # the same meta information as their changeset.
+ CheckInCS \
+ {All revisions have to agree with their revision changeset about the used meta information} \
+ {disagrees with its revision changeset @ about the meta information} {
+ SELECT CT.name, C.cid, F.name, R.rev
+ FROM changeset C, cstype CT, revision R, file F, csrevision CR
+ WHERE C.type = 0 -- revision changesets only
+ AND C.cid = CR.cid -- changeset --> its revisions
+ AND R.rid = CR.rid -- look at them
+ AND R.mid != C.src -- Only those which disagree with changeset about the meta
+ AND R.fid = F.fid -- get file of the revision
+ AND CT.tid = C.type -- get changeset type, for labeling
+ }
+ return
}
proc SymbolChangesets {} {
+ # This code performs a number of paranoid checks of the
+ # database, searching for inconsistent changeset/revision
+ # information.
+
+ return ; # Disabled for now, bottlenecks ...
+
+ upvar 1 n n ; # Counter for the checks (we print an id before
+ # the main label).
+
+ # The next two checks are BOTTLENECKS. In essence we are
+ # checking each symbol changeset one by one.
+
+ # TODO: Try to rephrase the checks to make more use of
+ # indices, set and stream operations.
+
+ # All revisions used by tag symbol changesets have to have the
+ # changeset's tag associated with them.
+ CheckInCS \
+ {All revisions used by tag symbol changesets have to have the changeset's tag attached to them} \
+ {does not have the tag of its symbol changeset @ attached to it} {
+ SELECT CT.name, C.cid, F.name, R.rev
+ FROM changeset C, cstype CT, revision R, file F, csrevision CR, tag T
+ WHERE C.type = 1 -- symbol changesets only
+ AND C.src = T.sid -- tag only, linked by symbol id
+ AND C.cid = CR.cid -- changeset --> its revisions
+ AND R.rid = CR.rid -- look at the revisions
+ -- and look for the tag among the attached ones.
+ AND T.sid NOT IN (SELECT TB.sid
+ FROM tag TB
+ WHERE TB.rev = R.rid)
+ AND R.fid = F.fid -- get file of revision
+ }
+
+ # All revisions used by branch symbol changesets have to have
+ # the changeset's branch associated with them.
+
+ CheckInCS \
+ {All revisions used by branch symbol changesets have to have the changeset's branch attached to them} \
+ {does not have the branch of its symbol changeset @ attached to it} {
+ SELECT CT.name, C.cid, F.name, R.rev, C.cid
+ FROM changeset C, cstype CT, revision R, file F, csrevision CR, branch B
+ WHERE C.type = 1 -- symbol changesets only
+ AND C.src = B.sid -- branches only
+ AND C.cid = CR.cid -- changeset --> its revisions
+ AND R.rid = CR.rid -- look at the revisions
+ -- and look for the branch among the attached ones.
+ AND B.sid NOT IN (SELECT BB.sid
+ FROM branch BB
+ WHERE BB.root = R.rid)
+ AND R.fid = F.fid -- get file of revision
+ }
+
+ # TODO
+ # The state has to contain at least one tag symbol changeset
+ # for all known tags.
+
+ # TODO
+ # The state has to contain at least one branch symbol changeset
+ # for all known branches.
+ return
}
proc Check {header label sql} {
upvar 1 n n
set ok 1
foreach {fname revnr} [state run $sql] {
set ok 0
trouble fatal "$fname <$revnr> $label"
+ }
+ log write 5 integrity "\[[format %02d [incr n]]\] [expr {$ok ? "Ok " : "Failed"}] ... $header"
+ return
+ }
+
+ proc CheckCS {header label sql} {
+ upvar 1 n n
+ set ok 1
+ foreach {ctype cid} [state run $sql] {
+ set ok 0
+ trouble fatal "<$ctype $cid> $label"
+ }
+ log write 5 integrity "\[[format %02d [incr n]]\] [expr {$ok ? "Ok " : "Failed"}] ... $header"
+ return
+ }
+
+ proc CheckInCS {header label sql} {
+ upvar 1 n n
+ set ok 1
+ foreach {cstype csid fname revnr} [state run $sql] {
+ set ok 0
+ set b "<$cstype $csid>"
+ trouble fatal "$fname <$revnr> [string map [list @ $b] $label]"
}
log write 5 integrity "\[[format %02d [incr n]]\] [expr {$ok ? "Ok " : "Failed"}] ... $header"
return
}
Modified tools/cvs2fossil/lib/c2f_pbreakacycle.tcl from [949289215b] to [642a0a9f2e].
@@ -76,14 +76,14 @@
cyclebreaker breakcmd [myproc BreakCycle]
state transaction {
LoadCommitOrder
cyclebreaker run break-all [myproc Changesets]
-
- repository printcsetstatistics
- integrity changesets
- }
+ }
+
+ repository printcsetstatistics
+ integrity changesets
return
}
typemethod discard {} {
# Pass manager interface. Executed for all passes after the
Modified tools/cvs2fossil/lib/c2f_pbreakrcycle.tcl from [685d23fe9c] to [630bffc4ad].
@@ -66,14 +66,14 @@
cyclebreaker breakcmd {::vc::fossil::import::cvs::cyclebreaker break}
state transaction {
cyclebreaker run break-rev [myproc Changesets]
+ }
- repository printcsetstatistics
- integrity changesets
- }
+ repository printcsetstatistics
+ integrity changesets
return
}
typemethod discard {} {
# Pass manager interface. Executed for all passes after the
Modified tools/cvs2fossil/lib/c2f_pbreakscycle.tcl from [258b058f24] to [87397f0522].
@@ -65,14 +65,14 @@
cyclebreaker breakcmd {::vc::fossil::import::cvs::cyclebreaker break}
state transaction {
cyclebreaker run break-sym [myproc Changesets]
+ }
- repository printcsetstatistics
- integrity changesets
- }
+ repository printcsetstatistics
+ integrity changesets
return
}
typemethod discard {} {
# Pass manager interface. Executed for all passes after the
Modified tools/cvs2fossil/lib/c2f_pinitcsets.tcl from [47d9786663] to [3b39c5f0f7].
@@ -132,14 +132,14 @@
state transaction {
CreateRevisionChangesets ; # Group file revisions into csets.
BreakInternalDependencies ; # Split the csets based on internal conflicts.
CreateSymbolChangesets ; # Create csets for tags and branches.
PersistTheChangesets
-
- repository printcsetstatistics
- integrity changesets
}
+
+ repository printcsetstatistics
+ integrity changesets
return
}
typemethod discard {} {
# Pass manager interface. Executed for all passes after the