tikir [Wed, 12 Sep 2001 19:35:39 +0000 (19:35 +0000)]
added a method to BPatch for waitUntillStopped
tikir [Mon, 10 Sep 2001 04:02:48 +0000 (04:02 +0000)]
Bug cleared for compilation of paradyn in inst.C
tikir [Fri, 7 Sep 2001 21:15:48 +0000 (21:15 +0000)]
commit for api change in CFG class
tikir [Fri, 7 Sep 2001 21:15:07 +0000 (21:15 +0000)]
initial dyninstAPI commit for code coverage
tikir [Fri, 7 Sep 2001 21:14:04 +0000 (21:14 +0000)]
Initial dyninstAPI commit for code coverage
schendel [Fri, 7 Sep 2001 15:22:30 +0000 (15:22 +0000)]
fix bug where aggComponents weren't getting properly initialized - - - -
mark each metricDefinitionNode of the mdn graph that gets built up for a
metric if it has aggComponents that need to have their start time or
initial actual value initialized; check whether top level mdn's have any
children mdn's with aggComponents that need to be initialized when a
process continues; fix updateWithDeltaValue to handle unusual MT mdn graph
schendel [Fri, 7 Sep 2001 15:22:28 +0000 (15:22 +0000)]
fix bug where aggComponents weren't getting properly initialized - - - -
mark each metricDefinitionNode of the mdn graph that gets built up for a
metric if it has aggComponents that need to have their start time or
initial actual value initialized; check whether top level mdn's have any
children mdn's with aggComponents that need to be initialized when a
process continues
bernat [Wed, 5 Sep 2001 21:59:01 +0000 (21:59 +0000)]
Removed debugging message.
schendel [Wed, 5 Sep 2001 19:06:38 +0000 (19:06 +0000)]
fix bug where initial mdn values for a mdn were getting setup before the
mdn had been "okayed" to be sampled, ie. the aggComponents had been
installed. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
add member variable to indicate when sampling has been "okayed", that is
when the aggComponents have been installed on a mdn; print out mdn type in
debug trace lines
gurari [Tue, 4 Sep 2001 19:48:47 +0000 (19:48 +0000)]
This commit merges the AGG_LEV enumeration into the MDN_TYPE enumeration.
MDN_TYPE now indicates the type of a metricDefinitionNode for both
threaded and non-threaded paradyn. I removed the aggLevel member of
metricDefinition node and also got rid of several #ifdef's.
hollings [Tue, 4 Sep 2001 17:05:28 +0000 (17:05 +0000)]
Fixed an delete that is was not needed.
schendel [Tue, 4 Sep 2001 17:02:26 +0000 (17:02 +0000)]
remove initialization dependency problems between non-local static objects
by having baseBucketSize be initialized in accessor function
gurari [Fri, 31 Aug 2001 21:44:18 +0000 (21:44 +0000)]
This commit fixes a bug in function relocation pertaining to the creation
of new instPoints after a function has been relocated, on sparc. The
problem was that when new instPoints for the rewritten function were being
created, the constructor was checking the original function to find out
what instructions were located around the instrumentation point. (This
knowledge is used to determine how to instrument the point). Instead, the
constructor should have been examining the rewritten function.
I added a new constructor to the instPoint class with a different parameter
list that examines the instructions in the rewritten function. I also changed
the function newCallPoint() so that rather than generating a new instPoint
object, an instPoint object is created before newCallPoint is called and that
instPoint is then passed into the function as a parameter.
gurari [Fri, 31 Aug 2001 18:15:23 +0000 (18:15 +0000)]
This commit fixes a bug with the insertion of nops during function
relocation on sparc. The buffer was being accessed by newOffset + i, where
both newOffset and i were being incremented each iteration. Removed the +i.
bernat [Thu, 30 Aug 2001 21:41:00 +0000 (21:41 +0000)]
Added a error check.
bernat [Thu, 30 Aug 2001 21:30:22 +0000 (21:30 +0000)]
This breaks apart the behavior we use when searching for a function by
address. It's actually a bad solution, but it works _FOR NOW_.
When we determine the address of a function, there are multiple
possible answers. The first is the offset the function is at within a
library. The second, given a running process, is the function's
address in that process' memory space. The third is the (possibly
relocated) address of the function.
Here's the problem. The function pd_Function::getAddress is supposed
to return the third value, but defaults to the first if the function
is not relocated. The two different values should really be returned
by different functions.
We were having problems finding a function by address because of this
behavior. Multiple matches were being returned because a single
address matched one function's absolute address (2 above) and another
function's offset (1 above). The offset should never have been
consulted, realistically.
Fix: check by address type 2 above, and then by address 3. NEVER check
by address 1 (which, unfortunately, we still do).
hollings [Thu, 30 Aug 2001 19:10:47 +0000 (19:10 +0000)]
Removed code to use full paht name for name_.
Not sure how/why it got there.
beyerj [Thu, 30 Aug 2001 18:20:51 +0000 (18:20 +0000)]
adding test cases for Fortran
beyerj [Thu, 30 Aug 2001 18:17:13 +0000 (18:17 +0000)]
adding a test case for Fortran
beyerj [Thu, 30 Aug 2001 18:12:54 +0000 (18:12 +0000)]
adding support for Fortran
hollings [Thu, 30 Aug 2001 15:59:04 +0000 (15:59 +0000)]
Use stl standard names for deleting items form a list.
schendel [Thu, 30 Aug 2001 01:58:51 +0000 (01:58 +0000)]
removed unnecessary passing of current time into
sampleAggregator::aggregate()
schendel [Thu, 30 Aug 2001 01:58:49 +0000 (01:58 +0000)]
added documentation for the new aggregation code; removed unused code to
delay aggregation code
hollings [Wed, 29 Aug 2001 23:25:23 +0000 (23:25 +0000)]
Changes to work towards compatability with IBM's version of dyninst.
major visible change is that the BPatch_{struct,union,function,etc}
enum is now called BPatch_Data{....}.
Also added a bunch of #ifdef IBM_BPATCH_COMPAT to provide
implementations of IBM specific functions (some of which will
probably be moved to our version too).
schendel [Tue, 28 Aug 2001 19:39:38 +0000 (19:39 +0000)]
replace parens with underscores for COMPILE line since parens causing
labs g++ 2.95.3 wrapper to fail
chadd [Tue, 28 Aug 2001 18:07:11 +0000 (18:07 +0000)]
This commit fixes the problem loading libdyninstRT.dll by
paradynd on WinNT/Win2k.
M dyninstAPI/src/pdwinnt.C
M rtinst/src/RTinst.c
hollings [Tue, 28 Aug 2001 15:42:32 +0000 (15:42 +0000)]
Added the -xs flag to force debugging into into executable rather than
.o file.
hollings [Tue, 28 Aug 2001 15:28:55 +0000 (15:28 +0000)]
Changes to improving parsing of Fortran common blocks with the SunPro
compilers. Should also fix some issues realted to local variables with
this compiler.
Added setLanguage support for Elf object format.
schendel [Tue, 28 Aug 2001 02:50:46 +0000 (02:50 +0000)]
fix bug so don't try to set initial actual value for aggComponents
associated associated with a sampleAggregator parent of a shared COMP
(component) type mdn; - - - - - - - - - - - - - - - - - - - - - - - - -
added to the debug tracing
schendel [Tue, 28 Aug 2001 02:48:42 +0000 (02:48 +0000)]
fix bug so don't try to set initial actual value for aggComponents
associated associated with a sampleAggregator parent of a shared COMP
(component) type mdn; - - - - - - - - - - - - - - - - - - - - - - - - -
add member to indicate if initial actual value has been set; assert if
an initial actual value is attempted to be set if has already been set
schendel [Tue, 28 Aug 2001 02:48:41 +0000 (02:48 +0000)]
fix bug so don't try to set initial actual value for aggComponents
associated associated with a sampleAggregator parent of a shared COMP
(component) type mdn; - - - - - - - - - - - - - - - - - - - - - - - - -
gave each mdn ability to know it's initial actual value
schendel [Tue, 28 Aug 2001 02:48:39 +0000 (02:48 +0000)]
fix bug so don't try to set initial actual value for aggComponents
associated associated with a sampleAggregator parent of a shared COMP
(component) type mdn; - - - - - - - - - - - - - - - - - - - - - - - - -
made certain debug trace statements print out more
bernat [Mon, 27 Aug 2001 19:33:43 +0000 (19:33 +0000)]
Fixed up name resolution (findFunction -> findFuncByAddr)
schendel [Thu, 23 Aug 2001 14:54:22 +0000 (14:54 +0000)]
changed use of visi_DataValues so specific bucket values to be returned
are indicated;
schendel [Thu, 23 Aug 2001 14:53:26 +0000 (14:53 +0000)]
updated DEPENDS file somewhat to reflect include file changes
schendel [Thu, 23 Aug 2001 14:50:10 +0000 (14:50 +0000)]
the old aggregation code, which is has been rewritten
schendel [Thu, 23 Aug 2001 14:49:49 +0000 (14:49 +0000)]
make_Nan(), implemented now in pdRate.C, hopefully can go away soon with
conversion of the visis to use new time and sample value types
schendel [Thu, 23 Aug 2001 14:44:52 +0000 (14:44 +0000)]
initialize graph data to NaN;
fixed smoothing code so that NaN data values aren't taken into account in
smoothing, was causing initial graph points to be to small (and rise from
close to zero) when smoothed
schendel [Thu, 23 Aug 2001 14:44:51 +0000 (14:44 +0000)]
changed use of visi_DataValues so specific bucket values to be returned
are indicated;
reorganized in various places
schendel [Thu, 23 Aug 2001 14:44:50 +0000 (14:44 +0000)]
changed visi_DataValues so instead of all datagrid bucket values being
returned, only specified bucket values are returned;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
schendel [Thu, 23 Aug 2001 14:44:49 +0000 (14:44 +0000)]
store the initial actual value for the metric;
made some inline functions non-inline because of size;
changed Values function so instead of all datagrid bucket values being
returned, only specified bucket values are returned;
passing metric type into functions for metric type specific handling;
changed method to get data for all of the buckets;
schendel [Thu, 23 Aug 2001 14:44:48 +0000 (14:44 +0000)]
made some inline functions non-inline because of size;
adjusted datagrid cell Value function so it still returns the actual value
for SampledFunction type of metrics and the normalized value for
EventCounter metrics, the implementation needs to be changed since the
format of the histogram has been changed so that the change in sample
value (instead of the actual value) is stored for each bucket for
SampledFunction metrics;
uses the initial actual value for the metric
schendel [Thu, 23 Aug 2001 14:44:46 +0000 (14:44 +0000)]
changed visi_DataValues so instead of all datagrid bucket values being
returned, only specified bucket values are returned
schendel [Thu, 23 Aug 2001 14:44:42 +0000 (14:44 +0000)]
added explicit template instantiation for new aggregation code
schendel [Thu, 23 Aug 2001 14:44:41 +0000 (14:44 +0000)]
created a maxSample constant of type pdSample
schendel [Thu, 23 Aug 2001 14:44:40 +0000 (14:44 +0000)]
added NaN constant of type pdRate; initialize to NaN by default;
add ability to print out "NaN" in output operator for pdRate
schendel [Thu, 23 Aug 2001 14:44:38 +0000 (14:44 +0000)]
remove global field from histogram class;
first step in removal of metric style field;
fixed so flushUnsentBuckets sets the lastBinSent;
keep track of initial actual value so that current currentActualValue can
be calculated;
schendel [Thu, 23 Aug 2001 14:44:36 +0000 (14:44 +0000)]
the rewrite of the aggregation code
schendel [Thu, 23 Aug 2001 14:44:35 +0000 (14:44 +0000)]
initialize to NaN by default; created a maxSample constant of type pdSample
schendel [Thu, 23 Aug 2001 14:44:34 +0000 (14:44 +0000)]
added NaN constant of type pdRate; initialize to NaN by default
schendel [Thu, 23 Aug 2001 14:44:33 +0000 (14:44 +0000)]
changed aggregation classes and implementation and updated clients to use
new aggregation code;
added members to turn on and off the debug streams
schendel [Thu, 23 Aug 2001 14:44:32 +0000 (14:44 +0000)]
pulled metric style declaration out of hist.h into seperate file
schendel [Thu, 23 Aug 2001 14:44:31 +0000 (14:44 +0000)]
add cvs id line; first step in removal of metric style field;
remove global field from histogram class;
keep track of initial actual value so that current currentActualValue can
be calculated;
schendel [Thu, 23 Aug 2001 14:44:29 +0000 (14:44 +0000)]
aggregation defines which can be included without importing full
aggregation header file
schendel [Thu, 23 Aug 2001 14:44:28 +0000 (14:44 +0000)]
not using, but added in case someone needs in the future
schendel [Thu, 23 Aug 2001 14:44:24 +0000 (14:44 +0000)]
changed aggregation classes and implementation and updated clients to use
new aggregation code;
schendel [Thu, 23 Aug 2001 14:44:22 +0000 (14:44 +0000)]
removed redundant setting of the initial daemon start time
schendel [Thu, 23 Aug 2001 14:44:20 +0000 (14:44 +0000)]
changed aggregation classes and implementation and updated clients to use
new aggregation code;
set initial start time for metrics from continue callback;
first step in removal of metric style field;
reconcile MT and non-MT checking of top level mdn type;
added on ostream<< function for a metricDefinitionNode;
set initial start time for metrics;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
add ability to update aggregation intervals;
reorganize updateValue and aggregation of levels;
schendel [Thu, 23 Aug 2001 14:44:17 +0000 (14:44 +0000)]
changed aggregation classes and implementation and updated clients to use
new aggregation code;
reorganized internal metrics;
removed passing of node weight and internal node indicator when passing
sample data from daemon to nfront-end;
first step in removal of metric style field;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
changed internal metric's member metricDefinitonNode field to be a top
level mdn type, so will work with sendInitialActualValue;
reconcile MT and non-MT checking of top level mdn type;
clean up metric completion in regards to final metric samples being sent;
updated to use new Vector erase function;
removed compiler warning;
set initial start time for metrics from continue callback;
process::continueProc(); set initial start time for metrics;
added on ostream<< function for a metricDefinitionNode;
add ability to update aggregation intervals;
reorganize updateValue and aggregation of levels;
schendel [Thu, 23 Aug 2001 14:44:16 +0000 (14:44 +0000)]
removed cvs log lines
schendel [Thu, 23 Aug 2001 14:44:15 +0000 (14:44 +0000)]
first step in removal of metric style field; remove compiler warning;
changed aggregation classes and implementation and updated clients to use
new aggregation code;
schendel [Thu, 23 Aug 2001 14:44:14 +0000 (14:44 +0000)]
remove compiler warning
schendel [Thu, 23 Aug 2001 14:44:13 +0000 (14:44 +0000)]
removed cvs log lines;
created consistent metric update value functions;
changed aggregation classes and implementation and updated clients to use
new aggregation code;
first step in removal of metric style field;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
schendel [Thu, 23 Aug 2001 14:44:12 +0000 (14:44 +0000)]
created consistent metric update value functions;
reorganized in internal metrics
changed aggregation classes and implementation and updated clients to use
new aggregation code;
first step in removal of metric style field;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
schendel [Thu, 23 Aug 2001 14:44:10 +0000 (14:44 +0000)]
changed bucket_width internal metric to sampling_rate;
reorganized internal metrics; first step in removal of metric style field;
schendel [Thu, 23 Aug 2001 14:44:08 +0000 (14:44 +0000)]
changed bucket_width internal metric to sampling_rate;
reorganized internal metrics;
first step in removal of metric style field;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
schendel [Thu, 23 Aug 2001 14:44:06 +0000 (14:44 +0000)]
include additional header file
schendel [Thu, 23 Aug 2001 14:44:05 +0000 (14:44 +0000)]
remove compiler warning; include additional header file
schendel [Thu, 23 Aug 2001 14:44:04 +0000 (14:44 +0000)]
remove compiler warning; add ability to update aggregation intervals;
removed unnecessary code;
schendel [Thu, 23 Aug 2001 14:44:01 +0000 (14:44 +0000)]
first step in removal of metric style field; cleaned up in various ways
changed aggregation classes and implementation and updated clients to use
new aggregation code; removed unneeded friend function declarations;
created consistent metric update value functions
schendel [Thu, 23 Aug 2001 14:43:59 +0000 (14:43 +0000)]
added ability to register a callback to be called when calling
process::continueProc()
schendel [Thu, 23 Aug 2001 14:43:58 +0000 (14:43 +0000)]
updated DEPENDS file to reflect include file changes
schendel [Thu, 23 Aug 2001 14:43:55 +0000 (14:43 +0000)]
changed aggregation classes and implementation and updated clients to use
new aggregation code; removed unneeded friend function declarations;
deleted aggregation operator aggAvg
schendel [Thu, 23 Aug 2001 14:43:54 +0000 (14:43 +0000)]
fixed bug from my previous commit, where I was translating NaN values into
0 values when transferring samples between front-end and visis, now NaN
values are preserved;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value
schendel [Thu, 23 Aug 2001 14:43:47 +0000 (14:43 +0000)]
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
for sampleFunction metrics, reconstructing each buckets actual value by
keeping a running total (integrating) of the samples
schendel [Thu, 23 Aug 2001 14:43:46 +0000 (14:43 +0000)]
reorganized sample data histogram callback function;
remove global field from histogram class;
add intelligence to know what to use for an aggregation interval;
schendel [Thu, 23 Aug 2001 14:43:44 +0000 (14:43 +0000)]
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
schendel [Thu, 23 Aug 2001 14:43:42 +0000 (14:43 +0000)]
changed aggregation classes and implementation and updated clients to use
new aggregation code;
removed unneeded friend function declarations;
added ability for a metricInstance to know whether an initialActualValue
has been sent on a perfStream;
remove global field from histogram class;
reorganized sample data histogram callback function;
added intelligence to know what to use for an aggregation interval;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
reorganized sample data histogram callback function;
first step in removal of metric style field; update various static member
schendel [Thu, 23 Aug 2001 14:43:41 +0000 (14:43 +0000)]
remove global field from histogram class;
changed aggregation classes and implementation and updated clients to use
new aggregation code; reorganized sample data histogram callback function;
updated to use new Vector erase function;
added ability for a metricInstance to know whether an initialActualValue
has been sent on a perfStream; first step in removal of metric style
field;
cleaned up function names and implementation of daemon's sending the time
it started to the front-end;
schendel [Thu, 23 Aug 2001 14:43:40 +0000 (14:43 +0000)]
first step in removal of metric style field; update various static member
definitions;
remove global field from histogram class;
cleaned up function names and implementation of daemon's sending the time
it started to the front-end;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
schendel [Thu, 23 Aug 2001 14:43:38 +0000 (14:43 +0000)]
cleaned up function names and implementation of daemon's sending the time
it started to the front-end;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
cleaned up calculation of network delay time, now take multiple samples
of network delay time and use the average of these samples;
reorganized batchSampleDataCallback;
schendel [Thu, 23 Aug 2001 14:43:37 +0000 (14:43 +0000)]
cleaned up function names and implementation of daemon's sending the time
it started to the front-end;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value;
cleaned up calculation of network delay time, now take multiple samples
of network delay time and use the average of these samples;
reorganized batchSampleDataCallback;
removed passing of node weight and internal node indicator when passing
sample data from daemon to front-end;
changed aggregation classes and implementation and updated clients to use
new aggregation code;
schendel [Thu, 23 Aug 2001 14:43:35 +0000 (14:43 +0000)]
cleaned up function names and implementation of daemon's sending the time
it started to the front-end
schendel [Thu, 23 Aug 2001 14:43:34 +0000 (14:43 +0000)]
changed aggregation classes and implementation and updated clients to use
new aggregation code;
added ability for a metricInstance to know whether an initialActualValue
has been sent on a perfStream
schendel [Thu, 23 Aug 2001 14:43:32 +0000 (14:43 +0000)]
removed passing of node weight and internal node indicator when passing
sample data from daemon to front-end;
cleaned up function names and implementation of daemon's sending the time
it started to the front-end;
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value
schendel [Thu, 23 Aug 2001 14:43:31 +0000 (14:43 +0000)]
changed so initial actual value is sent in addition to the change in
sampling value data, which allows actual value at any point to be
reconstructed by integrating across change in sample value and adding
initial actual value
schendel [Thu, 23 Aug 2001 14:43:29 +0000 (14:43 +0000)]
fixed a bug in a previous commit which caused NT to not compile
schendel [Thu, 23 Aug 2001 14:43:28 +0000 (14:43 +0000)]
reorganized some paradyn specific code
schendel [Thu, 23 Aug 2001 14:43:24 +0000 (14:43 +0000)]
changed MACRO into an inline function; added ability to register a callback
to be called when calling process::continueProc()
schendel [Thu, 23 Aug 2001 14:43:22 +0000 (14:43 +0000)]
reorganized internal metrics; changed MACRO into an inline function;
removed compiler warning; created consistent metric update value functions;
added ability to register a callback to be called when calling
process::continueProc()
schendel [Thu, 23 Aug 2001 14:43:20 +0000 (14:43 +0000)]
changed MACRO into an inline function
schendel [Thu, 23 Aug 2001 14:43:18 +0000 (14:43 +0000)]
added ability to register a callback to be called when calling
process::continueProc()
schendel [Thu, 23 Aug 2001 14:43:17 +0000 (14:43 +0000)]
removed compiler warning; reorganized internal metrics
schendel [Thu, 23 Aug 2001 14:43:14 +0000 (14:43 +0000)]
removed compiler warning
schendel [Thu, 23 Aug 2001 14:43:12 +0000 (14:43 +0000)]
added members to turn on and off the debug streams
schendel [Thu, 23 Aug 2001 14:43:11 +0000 (14:43 +0000)]
removed compiler warning; added a member which removes (erase) an element
schendel [Thu, 23 Aug 2001 14:43:09 +0000 (14:43 +0000)]
made default initialization of timeLength to be in uninitialized state
added additional timeStamp constants
schendel [Thu, 23 Aug 2001 14:43:08 +0000 (14:43 +0000)]
added const version of index operator
schendel [Thu, 23 Aug 2001 14:36:02 +0000 (14:36 +0000)]
added const version of index operator
bernat [Wed, 22 Aug 2001 18:26:54 +0000 (18:26 +0000)]
You know, "for (int i; i < foo; i++)" isn't really a wise idea.
Sigh.