<?xml version="1.0" encoding="utf-8"?>
<feed xml:lang="en" xmlns="http://www.w3.org/2005/Atom"><title>Recent changes to bugs</title><link href="https://sourceforge.net/p/utool/bugs/" rel="alternate"/><link href="https://sourceforge.net/p/utool/bugs/feed.atom" rel="self"/><id>https://sourceforge.net/p/utool/bugs/</id><updated>2005-12-14T12:48:54Z</updated><subtitle>Recent changes to bugs</subtitle><entry><title>Dominances out of roots</title><link href="https://sourceforge.net/p/utool/bugs/11/" rel="alternate"/><published>2005-12-14T12:48:54Z</published><updated>2005-12-14T12:48:54Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.netb501b04fd17978ed6252c255a226f3139757e723</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;I'm not sure that the solver currently solves&lt;br /&gt;
constraints correctly in which both the root and a hole&lt;br /&gt;
of the same fragment dominate nodes in the same hn&lt;br /&gt;
connected component. In such a case, the sub-solved&lt;br /&gt;
form of the hncc should of course go below the hole --&lt;br /&gt;
but the current solver might also record it below the&lt;br /&gt;
"other" children. We need to check this and if the bug&lt;br /&gt;
is there, fix it asap.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Cleaner DLL building on Windows</title><link href="https://sourceforge.net/p/utool/bugs/10/" rel="alternate"/><published>2005-07-25T17:00:37Z</published><updated>2005-07-25T17:00:37Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.net5198845c399c744655a2cf15b563577f38fc5940</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;At the moment, utool builds really prettily on Linux:&lt;br /&gt;
Most of the material is put into libdomgraph.so, and&lt;br /&gt;
the utool executable is linked dynamically against&lt;br /&gt;
libdomgraph.so.&lt;/p&gt;
&lt;p&gt;Under Windows, we do build a libdomgraph.dll, but then&lt;br /&gt;
all object files are linked together statically _again_&lt;br /&gt;
to build utool.exe. Then when the Swig/Java integration&lt;br /&gt;
is built, all object files are linked together&lt;br /&gt;
statically _again_ to build DomgraphSwig.dll.&lt;/p&gt;
&lt;p&gt;This is not as nice as it could be. We should link&lt;br /&gt;
utool.exe and DomgraphSwig.dll dynamically against the&lt;br /&gt;
DLL just like under Linux.&lt;/p&gt;
&lt;p&gt;The main thing we'd need to do for this is to create an&lt;br /&gt;
import library libdomgraph.lib when we build&lt;br /&gt;
libdomgraph.dll. Then we could tell the linker to link&lt;br /&gt;
utool and DomgraphSwig.dll against libdomgraph.lib, and&lt;br /&gt;
everything would be fine.&lt;/p&gt;
&lt;p&gt;The Microsoft linker will automatically generate&lt;br /&gt;
libdomgraph.lib if we give it a libdomgraph.def&lt;br /&gt;
declaration file, in which we list all exported&lt;br /&gt;
symbols. The thing that needs to be done is either (a)&lt;br /&gt;
to find a way to extract all symbols to make them&lt;br /&gt;
exported, or (b) to create a *.def file manually, in&lt;br /&gt;
such a way that it will stay up-to-date when we change&lt;br /&gt;
stuff in libdomgraph. Possibly there's an option (c):&lt;br /&gt;
the MSVC compiler seems to support a "declspec"&lt;br /&gt;
notation that you can put in front of methods (and&lt;br /&gt;
#define it to the empty string for non-Windows&lt;br /&gt;
compilers), and that will tell the compiler that this&lt;br /&gt;
symbol should be exported.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>holesem-comsem output codec writes too few brackets</title><link href="https://sourceforge.net/p/utool/bugs/9/" rel="alternate"/><published>2005-06-20T13:56:27Z</published><updated>2005-06-20T13:56:27Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.netfa614b3fb416477b1b9cb4d355895864343090e4</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;It generates too few closing brackets at the end of the&lt;br /&gt;
output formula. Example: utool convert&lt;br /&gt;
examples/thatwould.clls -o thatwould.hs.pl, and check&lt;br /&gt;
thatwould.hs.pl.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Swig support under Cygwin</title><link href="https://sourceforge.net/p/utool/bugs/8/" rel="alternate"/><published>2005-06-05T15:51:10Z</published><updated>2005-06-05T15:51:10Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.net242182e79644fdeedf81aca7315bdcd307dd8087</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;We need to support Swig under Cygwin.&lt;/p&gt;
&lt;p&gt;Swig works out of the box on Cygwin, in principle. I&lt;br /&gt;
have also figured out how to compile a DLL from the&lt;br /&gt;
files Cygwin produces. But at this point, the Java test&lt;br /&gt;
program hangs in the call to System.loadLibrary. This&lt;br /&gt;
is possibly because the DLL is linked against the&lt;br /&gt;
dynamic Cygwin library and this isn't allowed -- but&lt;br /&gt;
more research is clearly necessary here (and then the&lt;br /&gt;
problem needs to be fixed asap).&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Dominances out of non-holes</title><link href="https://sourceforge.net/p/utool/bugs/7/" rel="alternate"/><published>2005-04-19T12:37:52Z</published><updated>2005-04-19T12:37:52Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.net6bcb39fe4a6555dc018b79bcafb6e7686cacce57</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Constraints of the form X:a &amp;amp; X \dom Y are&lt;br /&gt;
automatically unsatisfiable if X and Y are different&lt;br /&gt;
variables. The Bodirsky et al. paper calls constraints&lt;br /&gt;
of this form "non-well-formed". Utool happily solves them.&lt;/p&gt;
&lt;p&gt;Implement a check at the beginning of the solver that&lt;br /&gt;
checks whether a constraint is non-well-formed and&lt;br /&gt;
returns unsolvability in this case. Condition: In the&lt;br /&gt;
compact version of the graph, every upper end of  a&lt;br /&gt;
dominance edge tree-dominates a hole.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>leaf-labeledness</title><link href="https://sourceforge.net/p/utool/bugs/6/" rel="alternate"/><published>2005-04-08T14:31:41Z</published><updated>2005-04-08T14:31:41Z</updated><author><name>Stefan Thater</name><uri>https://sourceforge.net/u/thater/</uri></author><id>https://sourceforge.net873dee2e1fec6ec651801b982855acc60bd13421</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;utool does not check whether a constraint is leaf-labeled &lt;br /&gt;
or not. &lt;/p&gt;
&lt;p&gt;For the MRS- and Hole-Semantics codecs, utool should &lt;br /&gt;
check for leaf-labeledness as otherwise solutions are &lt;br /&gt;
produced which do not correspond to the configuartions &lt;br /&gt;
of the input description. A minimal example is &lt;/p&gt;
&lt;p&gt;[label(x f(y))]&lt;/p&gt;
&lt;p&gt;which has one solution (solved form) but no configuration.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Consistent exit codes</title><link href="https://sourceforge.net/p/utool/bugs/5/" rel="alternate"/><published>2005-04-08T10:18:07Z</published><updated>2005-04-08T10:18:07Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.nete4dc8c51302e3a3477f06e4b7eab61120865cc92</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Now that utool classify returns exit codes in the range&lt;br /&gt;
of 0-31, we need to move the exit codes that signal&lt;br /&gt;
errors into a range of their own.&lt;/p&gt;
&lt;p&gt;Implement the following convention: exit code of &amp;gt;=128&lt;br /&gt;
signals errors; exit code of &amp;lt;= 127 signals successful&lt;br /&gt;
termination.&lt;/p&gt;
&lt;p&gt;While we're at it, change the exit codes for "solvable"&lt;br /&gt;
and "solve" so 0 means "not solvable" and 1 means&lt;br /&gt;
"solvable".&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Memory leaks in input codecs</title><link href="https://sourceforge.net/p/utool/bugs/4/" rel="alternate"/><published>2005-03-31T15:35:58Z</published><updated>2005-03-31T15:35:58Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.net93956960b70fa49548315cb257acc2fd81b2f114</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Several input codecs allocate memory for node labels&lt;br /&gt;
and node and edge names (with strdup and new[]) which&lt;br /&gt;
is never freed. This is a memory leak, which is&lt;br /&gt;
tolerable for command-line use (only one constraint is&lt;br /&gt;
parsed per run of the program), but may become a bit of&lt;br /&gt;
a problem for the server mode.&lt;/p&gt;
&lt;p&gt;One idea for solving this is to make EdgeData,&lt;br /&gt;
NodeData, and NodeLabels allocate their own memory for&lt;br /&gt;
the names and labels. However, this leads to one of the&lt;br /&gt;
following kinds of behaviour: (a) The program hangs in&lt;br /&gt;
malloc(); (b) the program aborts with a segmentation&lt;br /&gt;
fault in a malloc(). This is not a segfault in free()&lt;br /&gt;
for unallocated memory; the free() goes through ok,&lt;br /&gt;
it's only the next malloc() or strdup() or new[] that&lt;br /&gt;
fails. The attached version of nodedata.cc illustrates&lt;br /&gt;
the problem.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Input codec error handling</title><link href="https://sourceforge.net/p/utool/bugs/3/" rel="alternate"/><published>2005-03-30T15:08:39Z</published><updated>2005-03-30T15:08:39Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.netae3f4dbcc94e02ec15f55a6d6d05211c27b99228</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;We need a general strategy for handling errors in the&lt;br /&gt;
input codecs -- in particular: (a) parsing errors; (b)&lt;br /&gt;
application of Holesem/MRS input codecs to non-nets.&lt;br /&gt;
Presumably the best way would be to throw an exception&lt;br /&gt;
that is caught in utool.cc.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Edges into non-roots</title><link href="https://sourceforge.net/p/utool/bugs/2/" rel="alternate"/><published>2005-03-23T17:24:51Z</published><updated>2005-03-23T17:24:51Z</updated><author><name>Alexander Koller</name><uri>https://sourceforge.net/u/akoller/</uri></author><id>https://sourceforge.net5e46a955a88005e0dfc9eb5e49d36b881265f506</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Utool computes too many solved forms if a dominance&lt;br /&gt;
edge goes into a non-root. A minimal example looks as&lt;br /&gt;
follows: [label(x1 f(x2)) label(y1 g(y2)) label(y2 a)&lt;br /&gt;
dom(x2 y2)].&lt;/p&gt;
&lt;p&gt;Fix: Move all dominance edges into non-roots to roots.&lt;/p&gt;&lt;/div&gt;</summary></entry></feed>