summaryrefslogtreecommitdiff
path: root/tools
AgeCommit message (Collapse)AuthorFilesLines
2009-03-23Do the legwork necessary to intern strings and structures that were parsed ↵Daniel Burrows2-21/+184
from the log file. This drastically cuts down on the memory usage of the viewer, and provides a nice speed boost too. Loading larger log files will be practical now. It would be nice if we could use a hashtable for the strings, but how to do that? Probably too much work for a little tool like this.
2009-03-22Don't make left square braces terminate versions, so we can parse "foo ↵Daniel Burrows1-1/+1
[UNINST]". This doesn't introduce an ambiguity: the left square brace can introduce a dependency whose source was fixed up, but in that case the version number will always be present and the dependency will be separated by whitespace.
2009-03-22Add a special case to render the minimum tier as "minimum".Daniel Burrows1-0/+1
2009-03-22Support parsing from-dep-source choices.Daniel Burrows1-4/+12
2009-03-22Adjust how successors are detected, to work reliably (with the new code).Daniel Burrows1-24/+7
The "Enqueuing" lines aren't paired with the "Trying" lines they refer to. Rather than fiddle around the edges, I decided to just change the log format to emit a new line when trying a dependency generates a new successor. That means that the search tree that's rendered is now correct.
2009-03-22Support parsing the solutions written by the trace code in my development tree.Daniel Burrows1-0/+1
2009-03-22Label nodes that don't represent a productive step so it's clear what type ↵Daniel Burrows1-2/+2
they are.
2009-03-22Be more robust when showing the text corresponding to a choice, and show ↵Daniel Burrows1-13/+27
text in more cases.
2009-03-22Get the resolver log visualizer into a halfway working state.Daniel Burrows2-105/+205
It more or less works now; probably not worth spending a lot of time tweaking it further.
2009-03-22Get the resolver visualizer more-or-less working.Daniel Burrows3-142/+707
It still doesn't display log fragments, but the major problems are more-or-less dealt with.
2009-03-21Display a minimal progress indicator for the load time tester.Daniel Burrows1-1/+12
2009-03-21Nicely pretty-print choices in the list display.Daniel Burrows1-2/+18
2009-03-21Don't bother including a condition that you update if the old progress bar ↵Daniel Burrows1-2/+1
fraction is "much smaller" than the new one.
2009-03-21Use widgetShowAll to display the main window, so all its subparts really appear.Daniel Burrows1-1/+1
2009-03-21Destroy the progress bar window after the file is loaded.Daniel Burrows1-0/+1
2009-03-21Actually add visible columns to the tree display, and make sure to set the ↵Daniel Burrows1-1/+4
model information after we create its columns.
2009-03-21Add a trivial executable that can be used to see how long it takes to load a ↵Daniel Burrows2-0/+21
log file.
2009-03-21Show the current progress as a percent on the load progress bar, and make ↵Daniel Burrows1-1/+4
the UI more responsive by dispatching *all* events on each iteration.
2009-03-21Fix the logic that decides when to update the progress bar.Daniel Burrows1-4/+6
Before, we were updating it on every line. With this change, log files are still slow to load, but they load in a somewhat reasonable amount of time.
2009-03-21Move "while" into a separate utility module.Daniel Burrows2-7/+9
2009-03-20Finish an initial implementation of code to render the main tree view and ↵Daniel Burrows5-50/+363
add Cabal build files.
2009-03-16Start working on a tool to parse and graphically display dependency resolver ↵Daniel Burrows7-0/+1132
traces. Currently only parsing works; the GUI is built in Glade, but still needs to be hooked up to the parse. We might want to extend the log format so that more details can be parsed out (e.g., information about soft deps is a little sketchy now). Note: the output of the log category aptitude.resolver.search is used as input to this tool, but you probably want to disable the category aptitude.resolver.search.tiers or you'll get a truly HUGE log file (tens of gigabytes when I tested it).