vestigia

commit 70a30ecfce7bf56a2deed0510e93d8aefbfab62e
parent 83a5e59a7681a797cd5315696d88f163c43dfbad
Author: Jared Tobin <jared@jtobin.io>
Date:   Fri, 30 Oct 2026 18:35:40 +0400

read: objects resolved to the files they name

A harness reports each object as its loader reached it, which for a
library linked by soname is the soname's link (libfoo.so.1), while
memcheck names a frame's object by the file the link leads to
(libfoo.so.1.2.3). Neither the path nor the file name matched, so
every site in such a library went without an offset, and a consumer
placing sites by offset could not place it. read resolves the
reported paths before locating frames; one that names nothing where
read runs is kept as reported.

Diffstat:
M
CHANGELOG
 | 
3
+++
M
app/Main.hs
 | 
5
+++--
M
lib/Vestigia/Trace.hs
 | 
20
++++++++++++++++++++
M
test/Main.hs
 | 
19
+++++++++++++++++++
M
vestigia.cabal
 | 
3
+++

5 files changed, 48 insertions(+), 2 deletions(-)
diff --git a/CHANGELOG b/CHANGELOG
@@ -18,6 +18,9 @@
   * `record` appends the guest's ISA to its dump.
   * `diffOf` takes a `contract`; the command finds LLVM's
     `llvm-objdump` through `$VESTIGIA_OBJDUMP`.
+  * `read` resolves the objects a harness reports to the files they
+    name, as memcheck names a frame's object, so a library reached by
+    its soname's link has its sites placed by offset.
 
 - 0.1.0 (2026-10-22)
   * Initial version, with two instruments over a harness that marks
diff --git a/app/Main.hs b/app/Main.hs
@@ -54,7 +54,8 @@ import Vestigia.Layout (fixLayout)
 import Vestigia.Memcheck (Memcheck, readMemcheck)
 import Vestigia.Symbol (symbolize)
 import Vestigia.Trace
-  (Taint(..), readObjects, sitesIn, trace, traceJson, traceSummary)
+  (Taint(..), readObjects, resolveObjects, sitesIn, trace, traceJson,
+   traceSummary)
 
 data Opts = Opts {
     oValgrind :: !(Maybe FilePath)
@@ -289,7 +290,7 @@ markedObjects p = do
         \no secret (vestigia_secret was never called on any bytes); \
         \its run says nothing")
       exitWith (ExitFailure 2)
-    Right os -> pure os
+    Right os -> resolveObjects os
 
 readXml :: FilePath -> IO Memcheck
 readXml xml = do
diff --git a/lib/Vestigia/Trace.hs b/lib/Vestigia/Trace.hs
@@ -33,16 +33,19 @@ module Vestigia.Trace (
   , trace
   , sitesIn
   , readObjects
+  , resolveObjects
   , objectOffset
   , traceJson
   , traceSummary
   , classText
   ) where
 
+import Control.Exception (IOException, try)
 import qualified Data.List as L
 import qualified Data.Map.Strict as M
 import qualified Data.Text as T
 import Numeric (readHex, showHex)
+import System.Directory (canonicalizePath, doesPathExist)
 
 import Vestigia.Json
 import Vestigia.Memcheck
@@ -189,6 +192,23 @@ readObjects = traverse line . filter (not . T.null . T.strip) . T.lines
 
 -- | A frame's address in its object, where the harness reported the
 --   object: by its path, else by its file name.
+-- | A run's objects, each path resolved to the file it names, as
+--   memcheck names a frame's object: a loader reaches a library by
+--   the link its soname names, memcheck by the file the link leads
+--   to. A path that names nothing here, as in a report made
+--   elsewhere, is kept as reported.
+resolveObjects :: [(Integer, T.Text)] -> IO [(Integer, T.Text)]
+resolveObjects = traverse resolve
+  where
+    resolve (b, p) = do
+      let path = T.unpack p
+      here <- doesPathExist path
+      if not here then pure (b, p) else do
+        r <- try (canonicalizePath path)
+        pure (b, either (keep p) T.pack r)
+    keep :: T.Text -> IOException -> T.Text
+    keep p _ = p
+
 objectOffset :: [(Integer, T.Text)] -> Frame -> Maybe Integer
 objectOffset os f = do
   obj <- frObj f
diff --git a/test/Main.hs b/test/Main.hs
@@ -9,6 +9,8 @@ import qualified Data.Map.Strict as M
 import qualified Data.Set as S
 import qualified Data.Text as T
 import qualified Data.Text.Encoding as TE
+import System.Directory
+import System.FilePath ((</>))
 import Test.Tasty
 import Test.Tasty.HUnit
 
@@ -108,6 +110,23 @@ traceTests = testGroup "trace"
         @?= Just 0x401000
       objectOffset os (f "/lib/libc.so.6" 0x7000) @?= Nothing
 
+  , testCase "an object reached by a link is named by its file" $ do
+      tmp <- getTemporaryDirectory
+      let dir = tmp </> "vestigia-resolve-objects"
+          file = dir </> "liba.so.1.2.3"
+          link = dir </> "liba.so.1"
+      removePathForcibly dir
+      createDirectoryIfMissing True dir
+      writeFile file ""
+      createFileLink "liba.so.1.2.3" link
+      real <- T.pack <$> canonicalizePath file
+      os <- resolveObjects
+        [ (0x4000000, T.pack link), (0, "/nowhere/bin/harness") ]
+      removePathForcibly dir
+      os @?= [ (0x4000000, real), (0, "/nowhere/bin/harness") ]
+      objectOffset os (Frame 0x4000010 (Just real) Nothing Nothing Nothing)
+        @?= Just 0x10
+
   , testCase "a site is kept to the objects named" $ do
       tr <- one <$> fixture "probe.xml"
       length (trSites (sitesIn [ "probe" ] tr)) @?= 2
diff --git a/vestigia.cabal b/vestigia.cabal
@@ -44,6 +44,7 @@ library
       base >= 4.9 && < 5
     , bytestring
     , containers
+    , directory
     , process
     , text
 
@@ -78,6 +79,8 @@ test-suite vestigia-tests
       base
     , bytestring
     , containers
+    , directory
+    , filepath
     , tasty
     , tasty-hunit
     , text