With ksh 2020.0.0+really93u+20120801-10 installed and ksh93u+m not yet
installed, I have the following issue (always reproducible):

# aptitude

1. I type 'U' to upgrade. I get

aptitude 0.8.13 @ cventin   #Broken: 1   Disk: -151 MB         DL: 123 MB/123 MB

[1(1)/...] Actions: 7 keeps
e: Examine  !: Apply  .: Next  ,: Previous

2. I type 'g' to perform the pending operations, keeping these
unrelated packages. This gives:

I've output debug information with the following two cases, quitting
after step 6:
  * '+' is used (before 'm');
  * '+' is not used.

The only differences with --log-level=debug (after filtering out
the prefix of each log line) except a PID difference in the name
of the temporary file:

@@ -543851,7 +543851,7 @@
 aptitude_resolver.cc:1762 DEBUG aptitude.resolver.scores - ** Score: +1 for gcc-sh4-linux-gnu:amd64 4:10.2.1-1 (aptitude::ProblemResolver::OptionalScore).
 aptitude_resolver.cc:1762 DEBUG aptitude.resolver.scores - ** Score: +1 for gcc-sh4-linux-gnu:i386 4:11.2.0--1 (aptitude::ProblemResolver::OptionalScore).
 aptitude_resolver.cc:1762 DEBUG aptitude.resolver.scores - ** Score: +1 for gcc-sh4-linux-gnu:i386 4:10.2.1-1 (aptitude::ProblemResolver::OptionalScore).
-resolver_manager.cc:581 DEBUG aptitude.resolver.thread - Resolver thread: got a new job { solution number = 0, max steps = 500000, continuation = 0x556ab550e820 }
+resolver_manager.cc:581 DEBUG aptitude.resolver.thread - Resolver thread: got a new job { solution number = 0, max steps = 500000, continuation = 0x5630a053c820 }
 problemresolver.h:4592 INFO aptitude.resolver.search - Starting a new search.
 problemresolver.h:4442 INFO aptitude.resolver.search - Examining step 0 (0 actions): {};T(nop)S-100
 promotion_set.h:1945 DEBUG aptitude.resolver.search.costs - Inserting (T(nop, advance: 10000): {Install(coq:amd64 8.12.0-3+b3)}) into the promotion set.
@@ -543968,7 +543968,6 @@
 apt.cc:329 INFO aptitude.apt.globals - Closing apt cache.
 apt.cc:382 DEBUG aptitude.apt.globals - Done closing the apt cache.
 apt.cc:398 INFO aptitude.apt.globals - Loading apt cache.
-aptcache.cc:2253 DEBUG aptitude.apt.cache - aptitudeDepCache::sweep(): Removing ksh93u+m:amd64: it is unused.
 apt.cc:483 DEBUG aptitude.apt.globals - Emitting cache_reloaded().
 download_queue.cc:1012 INFO aptitude.downloadQueue - Shutting down the download queue.
 download_queue.cc:1045 INFO aptitude.downloadQueue - The download queue is now shut down.

[...]

And with --log-level=trace, the first difference changes to:

@@ -2798559,7 +2798559,8 @@
 aptitude_resolver.cc:1762 DEBUG aptitude.resolver.scores - ** Score: +1 for gcc-sh4-linux-gnu:amd64 4:10.2.1-1 (aptitude::ProblemResolver::OptionalScore).
 aptitude_resolver.cc:1762 DEBUG aptitude.resolver.scores - ** Score: +1 for gcc-sh4-linux-gnu:i386 4:11.2.0--1 (aptitude::ProblemResolver::OptionalScore).
 aptitude_resolver.cc:1762 DEBUG aptitude.resolver.scores - ** Score: +1 for gcc-sh4-linux-gnu:i386 4:10.2.1-1 (aptitude::ProblemResolver::OptionalScore).
-resolver_manager.cc:581 DEBUG aptitude.resolver.thread - Resolver thread: got a new job { solution number = 0, max steps = 500000, continuation = 0x5588ba253ed0 }
+resolver_manager.cc:565 TRACE aptitude.resolver.thread - Resolver thread: waiting for a job.
+resolver_manager.cc:581 DEBUG aptitude.resolver.thread - Resolver thread: got a new job { solution number = 0, max steps = 500000, continuation = 0x55c4f59b2ed0 }
 problemresolver.h:4592 INFO aptitude.resolver.search - Starting a new search.
 problemresolver.h:3239 TRACE aptitude.resolver.search - Marking the dependency why3-coq:amd64 1.3.3-3 -> {coq:amd64 8.12.0-3+b3} as unresolved in step 0
 problemresolver.h:2634 TRACE aptitude.resolver.search - Adding the solver Install(coq:amd64 8.12.0-3+b3 <why3-coq:amd64 1.3.3-3 -> {coq:amd64 8.12.0-3+b3}>) with initial cost (nop)

Hi Vincent,

thanks for the detailed bug report with all the steps to reproduce.

Vincent Lefevre wrote:
[...]

Correct, I can reproduce it, also with just pressing "U" again instead
of quitting, i.e. it seems as if every action which writes down and
rereads the current state triggers this unexpected behaviour.

		Regards, Axel