Activity for Gordon Kindlmann
-
Gordon Kindlmann committed [r7241] on Code
syncing with source
-
API NEW: new field NrrdIoState->declineStdioOnTTY, which instructs nrrdLoad/nrrdSave to decline reading/writing to stdin/stdout if the given filename is - AND stdin/stdout is a terminal (as per isatty). Can be over-ridden by using filename -=. This is long-wanted functionality that was only half-addressed, on the command-line input side, via nrrdHestNrrdNoTTY
-
now clang-format'ed
-
added ability to recognize -= as the same stdin/stdout that - is recognized as
-
source sync
-
unfortunately a mix of (1) new formatting with a new version of blue, and (2) changing how #defines in header files are processed: from a list of strings defined, to: a dict mapping from name to value, and a modest effort to apply those substitutions prior to sending things to cffi.cdef()
-
syncing with sources
-
syncing with sources
-
limnCbfTVT now computes scale=0 TVT, and that is now passed back to fctx->cornerCB corner tester
-
syncing with source
-
main change: now can use a callback function as part of determining which vertices count as a corner
-
syncing with source
-
fixing problem with nan generation from scale>0 vertex pos == orig pos
-
syncing with source
-
more complete use of nrrdHestNrrdNoTTY instead of nrrdHestNrrd
-
clang format, and using nrrdHestNrrdNoTTY
-
clang format, and using nrrdHestNrrdNoTTY
-
reformatting/clarifying comment
-
clang format, and using nrrdHestNrrdNoTTY
-
biffMovef didnt free messages at source key -> memory leak
-
clarifying comment; trying to fix memory error
-
fixing uses of AIR_STRLEN_x plus clang-format
-
fixing uses of AIR_STRLEN_x plus clang-format
-
fixing uses of AIR_STRLEN_x plus clang-format
-
fixing uses of AIR_STRLEN_x plus clang-format
-
fixing uses of AIR_STRLEN_x plus clang-format
-
fixing uses of AIR_STRLEN_x plus clang-format
-
fixing uses of AIR_STRLEN_x plus clang-format
-
fixing use of AIR_STRLEN_SMALL
-
synching with source
-
it may actually be working ...
-
tweaks to new flexiblity in loi==hii, and in its testing
-
tweaks to new flexiblity in loi==hii, and in its testing
-
still debugging
-
recognizing that having one corner in path is possible
-
now actually accepting more that type double
-
usefully acting on float-type data
-
synching with Teem source
-
some final comments
-
added wacky path handling for 3=spanlen, and maybe done
-
-roll options for debugging loops, and now path printing is uniform
-
debugging scripts, maybe done?
-
test script tweaks
-
clarifyig comments
-
refactored Newton-restep code to be less dumb and ugly
-
improved the nrp (Newton-based ReParameterization) so that it never takes bad steps; it did that previously
-
renaming simple to punt (and removing incorrect statements of straight path), and some more refined debugging statements
-
segfault fixed, debugging continues
-
adding memory of recursion depth to limnCbfMulti
-
another test script
-
multi-fit may be working now
-
fixing compilation/valgrind issues
-
making more progress, but now debugging segfault
-
number of colummns set by terminal
-
isolated a case where single fit is bad
-
typo in verbose output
-
this illustrates a situation for which single fit needs fixing
-
now lpu cbfit gets some defaults from limnCbfCtxNew
-
another script for debugging; may have found a bug (weird dependence on number of points in synthetic data)
-
re-wrote idxLift and limnCbfTVT, and things may finally be working, but testing ongoing
-
testing continues
-
testing in progress
-
one way to test limnCbfSingle
-
for testing limnCbfTVT
-
data generation for for testing
-
fctx->distMaxIdx is now set correctly, hopefully, and that and lots of other things still need to be tested more. Added documentation to reflect new term lifted indices, which is the right word for something GLK has been struggling to put into precise words
-
code is all there, but TVT computations still need fixing, and all of this still needs debugging
-
fixing memory leak
-
forgot to add this after having added the JSF code
-
limnCBFitSingle seems to work on tests so far; more to come ...
-
may have limnCBFitSingle working; added --help support to test/lpu
-
more progress; limnCBFFindTVT seems to work now
-
added warning about how AIR_MOD can give bad results if args differ in sign-ed-ness, and other formatting tweaks
-
finally making some progress on cubic spline fitting; still very much in progress; committing work so far to facilitate testing on different platforms
-
synching with source
-
finally re-started progress on limnCBF functions
-
refecting reality that Teem v2 is going to be next release
-
fixing writing fumble in comment
-
synching with source
-
fixing typo in comment
-
happy new year
-
what the hell was that
-
removing test on nrrdTypeIsIntegral[nrrdTypeBlock] from nrrdSanity()
-
nrrdTypeIsIntegral[nrrdTypeBlock] is no longer true; this was asserted in ~2002 but never made sense
-
synching with source
-
more explicitly disallowing block type
-
changing URL for where GLK learned this trick
-
tweaked handling of too-small ioctl-produced columns
-
sync with sources
-
adding and using hestParmColumnsIoctl() to contain in a single function smarts for using ioctl() to get terminal width, for the sake of hest usage
-
variable rename to reflect what is described in the comment
-
sync with source
-
finished BUT NOT TESTED double-precision versions
-
sync with sources
-
starting BUT NOT YET TESTED Jenkins Small Fast JSF RNG
-
fixing and improving description and usage info
-
more handling of NaN
-
better wording of diagnostic message
-
sync with source
-
smplr->verbPixel never initialized
1 >