Log message:
revert wrong fix for py-scipy python 3.6 deprecation, fix properly

Log message:
(geography/py-obspy) Updated 1.1.1 to 1.2.2 (build fix, too)

1.2.2 (doi: 10.5281/zenodo.3921997)
===================================

Changes:
 - obspy.core:
   * Fix wrong values in Stats object after deepcopy or pickle of Stats object
     for edge cases (see #2601)
 - obspy.clients.fdsn:
   * EIDA routing client: fix an issue that leaded to a request of *all* EIDA
     data when requesting an invalid, out-of-epochs time window for a valid
     station (see #2611)
   * update RASPISHAKE URL mapping to use https
   * fix a bug of not handling HTTPException in mass_downloader (see #2606)
   * use the client's set timeout in service discovery, too (see #2656)
   * add URL mapping for FDSNWS at UIB-NORSAR (see #2659)
 - obspy.clients.filesystem:
   * sds: continue get_all_stations() even if encountering an invalid channel
     code (see #2636)
 - obspy.clients.neic:
   * Make client socket blocking (see #2617)
 - obspy.io.nordic:
   * Fixed a bug raising an exception when reading a nordic file with a non
     positive-definite covariance matrix (see #2593)
   * Allow 60.0 seconds in the second field of origin times, allow NaN in
     the magnitude field (see #2627)
 - obspy.io.seiscomp:
   * Fixed an issue where a response has zero zeros or poles (see #2633)

1.2.1 (doi: 10.5281/zenodo.3706479)
===================================

Changes:
 - fix an installation issue with pip and setuptools version 46 (see #2578)
 - fix response plots when providing `axes=...` with a numpy array of Axes
   instances (see #2579)

1.2.0 (doi: 10.5281/zenodo.3674646)
===================================

Work on this release was in parts and among others supported by the following
institutions/companies and grants (in alphabetical order):
 - Earthquake Commision of New Zealand (EQC), grant 18/753
 - École et Observatoire des Sciences de la Terre - Université de Strasbourg
 - ETH Zürich
 - Friedrich-Schiller-Universität Jena
 - Geoscience Australia
 - Incorporated Research Institutions for Seismology (IRIS), NSF (SAGE) award
   :: EAR-1851048
 - Institut de Physique du Globe de Strasbourg
 - Institut de Physique du Globe de Paris
 - Institutions for Seismology (IRIS) through the PASSCAL Instrument Center at
   New Mexico Tech. The facilities of the IRIS Consortium are supported by the
   National Science Foundation under Cooperative Agreement EAR-1261681 and the
   DOE National Nuclear Security Administration.
 - Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Etneo (Italy),
   Allegato B2 DPC-INGV 2012-2021 Task 10
 - Ludwig-Maximilians-Universität München
 - National Institute for Occupational Safety and Health
 - Royal Netherlands Meteorological Institute (KNMI)
 - School of Geography, Environment and Earth Sciences, Victoria University of
   Wellington
 - The European Union’s Horizon 2020 research and innovation programme under
   the ChEESE project, grant agreement No. 823844
 - The Royal Observatory of Belgium
 - U.S. Geological Survey

Changes:
 - obspy.core:
   * Inventory objects have been adapted to StationXML 1.1 for details on
     changes see #2510 and
     https://github.com/FDSN/StationXML/blob/master/Changes.md
   * Fixed import of custom plugins (see #2423)
   * Fixed "==" comparison for Stream and Trace which was very slow in \ 
case of
     traces with >1e6 samples (see #2377)
   * Added almost_equal method for Stream and Trace classes (see #2286).
   * Casting FDSN identifiers to strings upon setting in the stats dictionary
     (see #1997).
   * UTCDateTime objects will now always evaluate equal if their string
     representations are equal (see #2049).
   * UTCDateTime objects now issue depreciation warnings when setting any
     attributes outside of init, or comparing UTCDateTime objects with
     different precisions (see #2077).
   * UTCDateTime objects can now accept hour, minute, second, and microsecond
     values greater than their normal limits by setting the strict keyword
     argument to False (see #2232).
   * Fixed UTCDateTime(..., julday=366) for non-leap years. This was returning
     January 1st of the next year in case of non-leap years being used. Now it
     properly raises an out-of-bounds ValueError (see #2369)
   * When reading StationXML/SC3ML, make sure to properly read empty string
     fields as empty strings instead of "None" (see #2519 and #2527)
   * Better ISO8601 detection for UTCDateTime objects and UTCDateTime(...,
     iso8601=False) now completely disables ISO8601 handling (see #2447)
   * Added replace method to UTCDateTime class (see #2077).
   * Added remove method to Inventory class (see #2088).
   * Added id property to WaveformStreamID (see #2131).
   * Added __str__ and _repr_pretty_ method for Comment class (see #2115)
   * Added __eq__ to QuantityError so empty instances equal None (see #2185).
   * Reworked the event scoped resource identifiers for the event classes
     hopefully fixing all edge-cases (see #2091).
   * Added a hook to allow users to customize finding objects for
     resource_ids which are not found via the normal means (see #2279).
   * Calling Stream.write(...) on an empty stream will now raise an
     ObsPyException consistently across all I/O plugins (see #2201)
   * Stream.get_gaps() will now properly report gaps within Traces that
     have masked arrays (i.e. Traces that have been merged without a fill
     value, see #2299 and #2300).
   * Added copy method to Inventory class (see #2322).
   * The Response.recalculate_overall_sensitivity() method now accepts integers
     (see #2338, #2343).
   * Added wildcard and url support to read_inventory (see #2326).
   * Modified stream.get_gaps() to deal with overlaps correctly (see #1403)
   * Added option "label_epoch_dates" to \ 
Inventory/Network.plot_response() to
     optionally add channel epoch start/end dates to legend labels (see #2309)
   * Deprecated the convert_id_to_quakeml_uri, regenerate_uuid, and
     get_quakeml_uri methods of the ResourceIdentifier class (see #2303).
   * Added get_quakeml_uri_str and get_quakeml_id methods to the
     ResourceIdentifier class (see #2303).
   * New method to create response objects directly from poles and zeros (see
     #1962).
   * Added Stream.stack method (see #2440).
   * Added a component field to the Stats object which allows to get and set
     the last character of the SEED channel (see #2484).
   * Fixed a bug in Stream.plot(type='section', reftime=..., ...) that caused
     wrong relative start times of traces relative to given reftime (see #2493)
   * Fixed a Windows-specific path case issue in a helper function that returns
     a list of untracked files in the git repository (see #2296)
   * Fix a bug that was causing an exception being raised in `Response.plot()`
     because a float was being passed down to numpy.linspace as length of array
     (see #2533)
   * Geographic select of inventory/network/station (see #2515)
   * Select traces in Stream based on an Inventory (see #2531)
 - obspy.clients.fdsn:
   * Add new `_discover_services` boolean flag to the Client, which allows the
     Client to skip the initial services query at instantiation.  This can
     reduce the load on service providers, but skips checks against unsupported
     query parameters.
   * Adding more location codes to the default priority list in the mass
     downloader (see #2155, #2159).
   * The mass downloader now raises a warning if all channels from a station
     have been deselected due to the default location priorities setting. This
     is a pure usability improvement as it has been confusing users
     (see #2159).
   * Make sure that streams fetched via FDSN are properly trimmed to user
     requested times if data center serves additional data around the start/end
     (see #1887, #2298)
   * Fix a problem that could spam subprocesses that were not closed in routed
     clients (see #2342 and #2344)
   * Make it possible to use signed EIDA tokens and also skip token validation
     completely (see #2297)
   * Adding a mapping for RASPISHAKE
 - obspy.clients.filesystem.tsindex:
   * Add new Client & Indexer modules based on IRIS time series index (see
     #2206)
 - obspy.clients.iris:
   * Results of distaz method are now returned as native floats (see #2499)
 - obspy.clients.neic:
   * Properly use specified timeout value (see #2450)
 - obspy.clients.seedlink:
   * Add method "get_info()" to fetch information on what
     networks/stations/locations/channels are served by the seedlink server
     (see #2405)
   * "get_waveforms()" can now be used with '*' and '?' wildcards in \ 
any part
     of requested SEED ID, i.e. network, station, location and channel (see
     #2405)
 - obspy.clients.seishub:
   * Properly handle fetching poles and zeros in presence of multiple metadata
     files for a given station (see #2411)
 - obspy.geodetics:
   * New utility function `inside_geobounds()` to check whether an object is
     inside a geographic bound (see #2515)
 - obspy.imaging:
   * obspy-scan can now be used with wildcarded SEED IDs when specifying what
     to plot after scanning data (see #2227)
   * Fix a problem in Scanner when loading npz on Python3 that was written on
     Python2 (see #2413)
   * Fix an issue that could make small landmasses not get plotted in basemap
     plots (see #2471, #2477)
   * Fixed a bug in Stream.plot(type='section', reftime=..., ...) that caused
     wrong relative start times of traces relative to given reftime (see #2493)
 - obspy.io:
   * Added read support for receiver gather format v. 1.6 (see #2070)
   * Added read support for FOCMEC 'out' and 'lst' files (see #2156)
   * Added read support for HypoDD 'pha' files (see #2378)
 - obspy.io.arclink:
   * Accommodate change in SeisComP3 publicID delimiter from '#' to '/' in
     ArclinkXML (see #2552)
 - obspy.io.dmx:
   * Add read support for INGV's DMX format (see #2452)
 - obspy.io.gcf:
   * Fixes Python 3.8 compatibility of GCF reader. (see #2505)
 - obspy.io.mseed:
   * Fix a bug resulting in an infinite loop when trying to read a FullSEED
     file without any data records (see #2534 and #2535)
   * Add ability to write int64 data to mseed if it can safely be downcast
     to int32 data, otherwise raises ObsPyMSEEDError. (see #2356)
   * The recordanalyzer can now detect calibration blockettes 300, 310,
     and 320 (see #2370).
   * Can now write zero sampling-rate traces. (see #2488, 2509)
 - obspy.io.nordic:
   * Add ability to read and write focal mechanisms and moment tensor
     information. (see #1924)
   * Add explicit warnings regarding unsupported sections of Nordic files.
   * Fix mapping of magnitude-types between MS to S and Ms to s.
   * Output preferred origin when writing to Nordic format instead of using
     the first origin (see #2195)
   * Include high-accuracy phase-pick reading and writing - high-accuracy is
     now the default phase-writing format, a boolean flag `high_accuracy`
     has been added to turn this off. (see #2351 and #2348)
   * Allow long-phase names (both reading and writing) - longer than 4 char.
     (see #2351)
   * Include AIN as takeoff-angle when reading and writing nordic files
     (see #2404).
   * Add error ellipses and read high-accuracy hypocenter lines (see #2451)
   * Fix the incorrect handling of events missing pick evaluation information
     (see #2520)
 - obspy.io.reftek:
   * Implement reading reftek encodings '16' and '32' (uncompressed data,
     16/32bit integers, see #2058 and #2059)
 - obspy.io.rg16:
   * Implement module to read waveforms and headers from fcnt format (see #2265).
   * Fix reading when start+endtime are inside one data packet (see #2485).
 - obspy.io.sac:
   * Fix bug writing inventory with SOH channels to SACPZ (see #2200).
   * Blank-pad character header variables, as opposed to NUL-padding them, to
     comply with SAC behavior (see #2543).
 - obspy.io.segy:
   * Raise nicer error messages when header packing fails (see #2194, #2196).
   * Show nice error message when trying to write a trace with too many samples
     in it (see #2358, #1393)
 - obspy.io.seg2:
   * Handle data format code 3 trace data (#2022, #2385).
   * Improve parsing of free-form entries (#2385).
   * Fix non-native endian data loading (#2385).
 - obspy.io.seiscomp:
   * Adding support for SC3ML 0.10 (see #2024).
   * Update xsl to allow conversion of amplitude picks not associated with
     origins (see #2273).
   * Very large performance improvement reading large sc3ml inventory files by
     pre-indexing sensors, dataloggers and responses and reducing lxml calls
     (see #2296).
   * When reading StationXML/SC3ML, make sure to properly read empty string
     fields as empty strings instead of "None" (see #2519 and #2527)
 - obspy.io.sh:
   * Add read support for SeismicHandler EVT event files (see #2109)
 - obspy.io.shapefile:
   * Add possibility to add custom database columns when writing catalog or
     inventory objects to shapefile (see #2012 and #2305)
 - obspy.io.stationxml:
   * When reading StationXML/SC3ML, make sure to properly read empty string
     fields as empty strings instead of "None" (see #2519 and #2527)
   * Inventory objects have been adapted to StationXML 1.1 for details on
     changes see #2510 and
     https://github.com/FDSN/StationXML/blob/master/Changes.md
 - obspy.io.quakeml:
   * Allow writing invalid ids but raise a warning
     (see #2104, #2090, #2093, #1872).
   * Skip invalid enumeration values during reading but raise a warning.
     (see #2106, #2098, #2095)
   * Catalogs with empty event description objects can be round-tripped (see
     #2339, #2340).
   * Correctly handle QuakeML native namespaces (if they are not matching the
     document root's namespace) as custom namespaces and parse them into
     `.extra` (see #2466)
 - obspy.io.xseed:
   * Ability to parse SEED files with extra newlines between blockettes
     (see #2383)
 - obspy.signal.cross_correlation:
   * Add new `correlate_template()` function with 'full' normalization option,
     required for correlations in template-matching
     (see #2035 and #2042).
   * 'domain' parameter in correlate function is deprecated in favour of new
     'method' parameter to be consistent with recent SciPy versions
     (see #2042).
   * Add `correlate_stream_template()` and `correlation_detector()`
     functions to detect events based on template matching (see #2315)
 - obspy.signal.PPSD:
   * Changed numpy-based serialization as to not require pickling (see #2424).
   * Fixed exact trace cutting for PSD segments (see #2040).
   * Timestamp representations internally and in npz I/O were changed to use
     integer nanosecond POSIX timestamps to avoid any potential floating point
     inaccuracies and since this is also what UTCDateTime is based on nowadays
     (see #2045).
   * Fixed the check for new PSD slices whether they should be added or whether
     they would add unwanted duplicated data (see #2229).
   * Fix `period_lim` option when `xaxis_frequency=True` (see #2246).
   * Added `allow_pickle` parameter to `PPSD.add_npz` and `PPSD.load_npz` and
     set its default to `False` (see #2457).
   * Added representation of earthquake models from Clinton & Heaton (2002) in
     'plot()' method using option 'show_earthquakes' (see #2455).
 - obspy.signal.polarization:
   * Fix an issue with covariance matrix in vidale algorithm and make adaptive
     windowing opt-out (see #2565)
   * Fix an issue in selecting Z/N/E traces from given stream (see #2365)
 - obspy.signal.trigger:
   * Fix a bug in AR picker (see #2157)
   * Option to return Baer-Kradolfer characteristic function from pk_mbaer
     function added (see #2341)
 - obspy.taup:
   * Fix cycling through colors in ray path plots (see #2470, #2478)
   * Fix a floating point issue on IBM machines (see #2559, #2560)

Log message:
math/blas, math/lapack: Install interchangeable BLAS system

Install the new interchangeable BLAS system created by Thomas Orgis,
currently supporting Netlib BLAS/LAPACK, OpenBLAS, cblas, lapacke, and
Apple's Accelerate.framework.  This system allows the user to select any
BLAS implementation without modifying packages or using package options, by
setting PKGSRC_BLAS_TYPES in mk.conf. See mk/blas.buildlink3.mk for details.

This commit should not alter behavior of existing packages as the system
defaults to Netlib BLAS/LAPACK, which until now has been the only supported
implementation.

Details:

Add new mk/blas.buildlink3.mk for inclusion in dependent packages
Install compatible Netlib math/blas and math/lapack packages
Update math/blas and math/lapack MAINTAINER approved by adam@
OpenBLAS, cblas, and lapacke will follow in separate commits
Update direct dependents to use mk/blas.buildlink3.mk
Perform recursive revbump

Log message:
py-obspy: fix PLIST

Log message:
py-obspy: Python 2.7 is not supported any more because of py-scipy

Log message:
(geography/py-obspy) PLIST variation for py 27 36 37

Log message:
Update to 1.1.1

Upstream changes:
1.1.1: (doi: 10.5281/zenodo.1040770)

    General:
        Tests pass with numpy 1.14 (see #2044).
        Map plots now also work with matplotlib >= 2.2 (see #2089).
    obspy.core:
        UTCDateTime now raises a meaningful exceptions when passing invalid or
        out-of-bounds 'julday' during initialization (see #1988)
        Fix pickling of traces with a sampling rate of 0 (see #1990)
        read_inventory() used with non-existing file path (e.g. typo in filename)
        now shows a proper "No such file or directory" error message \ 
(see #2062)
        Fix Trace.times(type='matplotlib') being slow (see #2112)
        read_events() and read_inventory() now trial most common plugins first
        (QuakeML/StationXML, ...) in case of automatic file format detection (i.e.
        when file type was not explicitly specified, see #2113)
        Event instances with Origin instances that have do not have defined
        latitude/longitude attributes will no longer raise a TypeError when
        creating a string representation (see #2119 and #2127).
        Fix Stream.get_gaps() when a trace is completely overlapping another trace
        (see #2218).
        Fix Exception when comparing ComparingObjects (see #2220).
        Fix UTCDateTime.strftime() when year is <1900 on Python 2 (see #2167)
        Inventory objects are more convenient to create now. Network, station, and
        channel codes can now be optional. Additionally the source parameter of
        inventories must no longer be specified at init time (see #2307, #2314).
    obspy.clients.arclink:
        Raise a warning at import time that the ArcLink protocol will be
        deprecated soon (see #1987).
    obspy.clients.fdsn:
        Mass downloader: Priority lists are now correctly overwritten if channel
        and/or location are set (see #1810, #2031, #2047).
        A few fixes and stability improvements for the mass downloader (see
        #2081).
        Fixed routing startup error when running under certain locales (see #2147)
        Update the IPGP mapping (see #2268).
        Adding a mapping for the KNMI (see #2270) services.
    obspy.clients.nrl:
        Set input units of overall sensitivity to input units of first stage in
        NRL.get_response() (see #2248)
    obspy.geodetics:
        Fix the vincenty inverse calculation for equatorial lines (see #2282).
    obspy.imaging:
        Normalize moment tensors prior to plotting in the mopad wrapper to
        stabilize the algorithm (see #2114, #2125).
        fix some map plotting issues with cartopy and local projection (see #2193,
        #2204)
    obspy.io.ascii:
        Fixes an issue with the time representation (see #2165, #2179).
    obspy.io.cnv:
        Bugfix when phase_mapping is passed as argument when writing a Catalog
        object to CNV (see #2001)
    obspy.io.css:
        Fix automatic filetype detection (see #2160 and #2162)
    obspy.io.gcf:
        Fix reading stream ID for station/channel code in header (see #2289,
        #2311)
        Fix bitmask in getting compression code (see #2290, #2310)
    obspy.io.mseed:
        Ability to read files that have embedded chunks of non SEED data. (see
        #1981, #2057).
        Fix util.get_start_and_end_time returning sample rate = 0 when sample rate
        = 1 (see #2069)
        Avoid showing invalid warnings when guessing endian during parsing
        timestamps (see #1988)
        util.get_record_information() now works correctly for negative sampling
        rate factors and multipliers (see #2030, #2191).
    obspy.io.nordic
        Bug-fix for amplitudes without magnitude_hint (see #2021)
        Bug-fix for wavefiles with full path stripping (see #2021)
        Bug-fix for longitudes between -100 and -180 (see #2197)
    obspy.io.reftek:
        Fix problems reading some Reftek 130 files, presumably due to floating
        point accuracy issues in comparing timestamps. Internal representation of
        time stamps is changed to integer nanosecond POSIX timestamp (see #2036,
        #2038, #2105)
        Fix a bug that prevents reading files that have no data in first channel
        (see #2101)
    obspy.io.sac:
        Allow passing on the byteorder flag from the top-level obspy.read()
        function (see #2285, #2292).
    obspy.io.seiscomp:
        Fix inventory read when maxClockDrift is unset in SC3ML (see #1993)
        Fix the reading of FIR coefficients when multiple whitespaces in SC3ML
        (see #2259)
        Fix the reading of the poles and zeros when multiple whitespaces in SC3ML
        (see #2260).
        Fix reading files with zero sampling rates (see #2294 and #2293)
        Fix divide by zero error when parsing sc3ml files of zero sampling rage
        (see #2294).
    obspy.io.stationxml
        Allow writing of dates before 1900 also on Python 2 (see #2013, #2015).
        Write the UTC time zone specifier to all times (see #2015).
        Units of first response stage as well as unit response stages are now
        determined with some heuristics (see #2250, #2318).
    obspy.io.xseed:
        Third condition to split blockettes when reading RESP files. Now more
        forgiving for slightly different files (see #2170, #2189)
    obspy.signal:
        Allow singular COUNT units in evalresp (see #2003, #2011).
        Fix an evalresp issue in case of an analog PAZ stage zero denominator (see
        #2171 and #2190)
        PPSD: for safety reasons, raise an ObsPyException if trying to read a PPSD
        npz file that was written with a newer version of the npz representation
        than is used by current ObsPy version (see #2051)
        The ar_pick() trigger function now raises an error if the three data
        arrays don't have the same length (see #1801, #2148).
        fix a precision issue in AR picker in case of low amplitude input (see
        #2252 and #2253)
    obspy.taup:
        Fallback to linear slowness interpolation for very small and shallow
        layers (see #2126, #2129).
        Fix bug preventing constant-velocity models with discontinuities at every
        layer boundary from being built (see #2264).
        More robust resize method so TauPy now works properly on Python 3.7 (see
        #2280, #2319).

Log message:
Now DEPENDS on py-matplotlib rather than buildlinking

Log message:
Restore plist vars, clobbered by last update.

Log message:
Update to 1.1.0

Upstream changes:
1.1.0: (doi: 10.5281/zenodo.165135)
 - General:
   * Read support for Guralp Compressed Format (GCF) waveform data,
     obspy.io.gcf (see #1449)
   * Read support for Reftek 130 (rt130) waveform data,
     obspy.io.reftek (see #1433)
   * Add Nordic format (s-file) read/write (see #1517)
   * Read and write support for events in the SCARDEC catlogue format
     (see #1391).
   * Read support for IASPEI ISF ISM 1.0 Bulletin event data,
     (see #1946)
   * Write support for AH (Ad Hoc version 1) format (see #1754)
   * Client to access the Nominal Response Library (NRL) (see #1185).
   * `obspy.read_inventory()` can now read dataless SEED and RESP files
     (see #1185).
   * change version number scheme for scenarios when no official version number
     can be determined (see #1889 and #1916)
   * Support for the IRIS Federator and EIDAWS FDSNWS web routing services
     (see #1779 and #1919).
 - obspy.core:
   * UTCDateTime is now based on nanoseconds (long) instead of a unix
     timestamp in microseconds (float) - resulting in higher precision and
     support for years 1-9999 (see #1325)
   * Ensure that Trace.data is always C-contiguous in memory (see #1732)
   * Event/ResourceIdentifier is now object aware, meaning even if two
     objects share a resource_id the distinct objects will be returned with
     the get_referred_object method provided both are still in scope. If one
     of the objects gets garbage collected, however, a warning will be issued
     and the behavior will be the same as before (see #1644).
   * Better error message when attempting to write invalid QuakeML resource
     ids (see #1699).
   * Stream/Trace.write() can now autodetect file format from file extension
     (see #1321).
   * New convenience property `.matplotlib_date` for `UTCDateTime` objects to
     get matplotlib datetime float representation (which can be used in
     time-based matplotlib axes, e.g. by Stream.plot(); see #1339).
   * Trace.times() has new options `type` and `reftime` to support fetching an
     array of sampletimes in various different timing varieties \ 
("relative":
     the old default, float relative to trace starttime or `reftime` in
     seconds; "utcdatetime": absolute times as UTCDateTime objects;
     "timestamp": array of float POSIX timestamps, compare
     `UTCDateTime.timestamp`; "matplotlib": array of float matplotlib \ 
dates,
     useful for plotting on matplotlib time axes; see #1307)
   * A trace's stats.network/station/location/channel can now also be set in
     one line using a SEED ID string (e.g. `trace.id = "GR.FUR..HHZ"`,
     see #1439).
   * Instrument correction for response list stages originating from inventory
     objects (see #1514).
   * `Stream.rotate(...)` can now also be used to rotate unaligned channels to
     Z-N-E, given an Inventory (see #1310)
   * Non finite floats (NaN, inf, -inf) can now no longer be set for all
     event objects (see #1597).
   * Instrument responses can now also be calculated for a given list of
     frequencies (see #1598).
   * Order of extra tags for event type classes serialized to QuakeML can now
     be controlled by using an OrderedDict (see #1617)
   * Bode plots can now optionally plot the phase in degrees (see #1763).
   * `Stream.select()` now also works on the component level if channels only
     have one letter (see #1847).
   * Now strips all invalid characters from the temporary filenames used for
     downloading data using the `read_X()` methods (see #1958).
 - obspy.clients.earthworm:
   * Much faster trace unpacking (see #1762).
 - obspy.clients.fdsn:
   * empty SEED codes (e.g. ``network=''``) will now be properly sent to the
     server as options and not omitted, which led to wildcard matching (for
     details see #1578)
   * The mass downloader now has `exclude_networks` and `exclude_stations`
     arguments to not download certain pieces of data. (see #1305)
   * The mass downloader can now download stations that are part of a given
     inventory object.
   * The mass downloader now also works with restricted data. (See #1350)
   * No data (HTTP 204) responses now raise `FDSNNoDataException` rather than
     the more general `FDSNException`.
   * Fixing cross implementation of bulk waveform and station requests (see
     #1685).
   * Adding mappings for the TEXNET (see #1852) and the ICGC (see #1902)
     services.
   * Support for the non-standard EIDA token authentication (see #1928).
 - obspy.imaging:
   * The functionality behind the `obspy-scan` command line script has been
     refactored into a `Scanner` class so that it can be reused in custom
     workflows. (see #1444)
 - obspy.imaging.cm:
   * new colormap: viridis_white. This is a modification of viridis that
     goes to white instead of yellow but remains perceptually uniform. It
     is especially useful for printing when an image should merge with the
     white background.
 - obspy.imaging.waveform:
   * Support for filling the wiggles when plotting sections (horizontal and
     vertical, see #1445).
 - obspy.io.arclink:
    * Read support for Arclink Inventory XML (see #1539)
    * default for `route` parameter in metadata requests is changed to `False`
      (see #1756)
 - obspy.io.ascii:
    * Custom formatting of sample values when writing SLIST and TSPAIR.
 - obspy.io.datamark:
    * Renamed without deprectation to obspy.io.win to match its original name.
      Datamark is a datalogger, saving the WIN format.
 - obspy.io.gse2:
    * Read support for GSE2.0 bulletin (see #1528)
 - obspy.io.nlloc:
    * Also parse author information and COMMENT line (see #1484)
    * Fix reading hypocenter files created by NonLinLoc versions of the 6.0.x
      beta branch (see #1760 and #1783)
 - obspy.io.quakeml:
    * Read and write support for nested custom tags (see #1463)
    * Fix some minor bugs that could lead to empty stub elements, e.g. like
      empty MomentTensor when reading and later writing again a QuakeML file
      with a FocalMechanism but no MomentTensor, potentially resulting in
      QuakeML files that breach the QuakeML schema (see #1896)
 - obspy.io.seiscomp:
    * Read and write support for SC3ML event (see #1638 and #1848)
    * Fix bug where files with arbitrary publicIDs and files with missing
      depth, latitude, longitude, or elevation tags could not be read
      (see #1817)
 - obspy.io.stationtxt:
    * Write support for stationtxt format (see #1466)
 - obspy.io.stationxml:
    * Read and write support for custom tags (see #1024)
    * No longer add the (unused) time zone field to StationXML datetimes to
      follow the example of big data centers. (see #1572)
    * Level of detail can be specified during inventory write (see #1830)
      using the level keyword (one of: network, station, channel, response).
    * Skip empty and incomplete channels during reading (see #1839, #1840).
 - obspy.io.segy:
    * Fixing an issue when comparing two still packed SEG-Y trace headers
      (see #1735).
    * Iterative reading of large SEG-Y and SU files with
      `obspy.io.segy.segy.iread_segy` and `obspy.io.segy.segy.iread_su`.
      (see #1400).
    * Write correct revision number (see #1737).
    * Textual headers will now always contain the file revision number and the
      end header mark if nothing else exists at these positions (see #1738).
    * The SEG-Y format detection now also checks the format version number
      (see #1781).
    * Enable reading SEG-Y files that have day of year 0 in trace header
      (see #1722).
    * Write textual file headers also if given as a text string
      (see #1811, #1813).
 - obspy.io.css:
   * Read support for NNSA KB Core format waveform data. (see #1332)
 - obspy.io.mseed:
   * New generic get_flags() utility function able to retrieve statistics
     about all fixed header flags and the timing quality. This makes the
     get_timing_and_data_quality() function obsolete which is thus
     deprecated and will be removed with the next release. The get_flags()
     function is also much faster. (see #1141)
   * Always hook up the libmseed logging to its Python counterpart to avoid
     some rare segfaults. (see #1658)
   * Update to libmseed v2.19.5 (see #1703, #1780, #1939).
   * Correctly read MiniSEED files with a data offset of 48 bytes (see #1540).
   * InternalMSEEDReadingError now called InternalMSEEDError and
     InternalMSEEDReadingWarning now called InternalMSEEDWarning as both
     can now also be raised in non-reading contexts (see #1658).
   * Should no-longer segfault with arbitrarily truncated files (see #1728).
   * Will now raise an exception when attempting to directly read mini-SEED
     files larger than 2048 MiB (#1746).
   * `.stats.mseed` attributes are no longer per-file but per-trace where
     applicable (see #1782).
   * `get_record_information()` - Don't fail if the word order is invalid.
 - obspy.io.nlloc:
   * Set preferred origin of event (see #1570)
 - obspy.io.nordic:
   * Add Nordic format (s-file) read/write (see #1517)
 - obspy.io.win:
   * see obspy.io.datamark.
 - obspy.io.xseed:
   * Added azimuth and dip to the get_coordinates() function. (see #1315)
   * Fixing some issues with the get_resp() output on Python 3 (see #1748).
   * Can now also parse RESP files (see #1185).
   * Can transform responses in the Parser object to ObsPy Inventory objects
     (see #1185).
 - obspy.scripts:
   * obspy-scan command line script now also plots and prints overlaps
     alongside gaps (see #1366)
   * obspy-plot now has option to disable min/max plot (see #1583)
 - obspy.signal:
   * fixed a bug in calibration.rel_calib_stack (resulting amplitude response
     had wrong scaling if using non-default "overlap_fraction", see #1821)
   * fixed a bug in coincidence_trigger() with event templates. when a template
     with mismatching SEED ID was encountered all following (potentially valid)
     templates were skipped as well (see #1850)
   * New obspy.signal.quality_control module to compute quality metrics from
     MiniSEED files. (see #1141)
   * New correlate function for calculating the cross-correlation function
     (new implementation based on Scipy).
     To calculate the shift of the maximum of the cross correlation use
     xcorr_max. The old xcorr function is deprecated but currently still
     exists (see #1585).
   * New obspy.signal.regression module to compute linear regressions, with or
     without weights, with or without allowing for an intercept. (see #1716,
     #1747)
   * add new plotting capabilities to PPSD (temporal variations per frequency
     and spectrogram-like plot) and also make underlying processed PSDs
     available via `PPSD.psd_values` property (see #1327)
   * Fixed bug in `rotate2zne()` for non-orthogonal configurations
     (see #1913, #1927).
   * Fixed build warnings in evalresp, partially backported from evalresp
     4.0.6 (see #1939).
 - obspy.taup:
   * Add obspy.taup.taup_geo.calc_dist_azi, a function to return the distance,
     azimuth and backazimuth for a source - receiver pair. (see #1538)
   * Fixing calculations through very small regional models. (see #1761)
   * Updated ray path plot method, added travel time plot method, and wrapper
     functions for both ray path and travel time plotting. (see #1501, #1877)