| Age | Commit message (Collapse) | Author |
|---|
|
OTP-9794
Stopping configuration (throw/catch) at first error instead of
continuing through all steps and then returning the error at the
end.
Start of reltool_server will no longer succeed if the configuration
causes an error.
reltool:get_status can only return {error,Reason} in the case where
the reltool_server has terminated. In all other cases it will return
{ok,Warnings}.
Bug fix in this commit:
* warnings are no longer duplicated in pop-up or return from reltool_server
|
|
OTP-9794
Backup old configuration before starting re-configuration so rollback
is possible if configuration fails.
Store last configuration including derivates so undo does no longer
need to refresh and analyse everything from disk.
|
|
OTP-9968
Make sure that inlined applications in an escript is included/excluded
as the escript itself, and forbid explicit configuration of the
inlined application.
|
|
OTP-9792
Start of reltool GUI sometimes crashes with a badmatch in
reltool_sys_win:do_init/1 because the #sys record fetched with
reltool_server:get_sys/1 differs from the #sys record returned from
reltool_server:start_link/1. This has been
corrected. reltool_server:start_link/1 no longer retuns the #sys
record.
|
|
OTP-9794
OTP-9968
The main idea behind the data structure in reltool_server is that the
state shall reflect what is explicitly configured, and the tables
shall contain this configuration plus everything that is derived. In
some cases, however, this was not the complete truth:
* the application table was never read
* the module table was never updated on undo
* the state contained a lot more than what was explicitly configured
This commit re-writes major parts of the reltool_server for the sake
of unifying the way the state and tables are updated:
* The list of applications in the state now only contains those
applications and modules for which there are explicit settings in
the configuration (given at startup or changed from the GUI)
* When changing any bit of the configuration, the tables are always
emptied and every part is derived again from the configuration found
in the state
* All configuration changes now cause a re-read of the file system,
meaning that if something has changed in the file system it will be
reflected in the result of the configuration change. This is the
case even if no file system related configuration is changed
(e.g. root dir or lib dirs)
(*POSSIBLE INCOMPATIBILITY*)
* Requests for applications and modules from the GUI now always read
the tables, not the state
* When loading a new configuration file via the GUI, the old
configuration is completly scratched, and only the new is valid
(*POSSIBLE INCOMPATIBILITY*)
* The handling of escripts which include archives of applications is
changed to always produce one #app record for the escript in
addition to one for each inlined application. All modules are listed
as parts of the inlined application where it belongs and not as part
of the escript's #app record. This is a temporary solution which
will be modified and improved.
The following bugs are corrected by this commit:
* Loading a config which contains an escript via the GUI menu did not
produce the same #app record as when loading the same configuration
at reltool start. Paths, version and label could differ.
* Loading config with same escript (source) twice caused reltool to
add same module twice in #app.mods
* Loading config with same escript (inlined beam) twice caused reltool
to fail saying module is included by two different applications
* Loading config which in addition to an existing escript also adds
another escript for which the name sorts before the existing one
would cause reltool to fail saying "Application name clash"
|
|
OTP-9794
OTP-9968
The following test cases are added:
* create_standalone_beam
* create_standalone_app
* create_multiple_standalone
* load_config_escript_path
* load_config_same_escript_source
* load_config_same_escript_beam
* load_config_add_escript
Most of them are temporarily skipped since they re-produce known
problems that will be corrected in a later commit.
|
|
OTP-9794
Test cases create_release_sort and create_script_sort are added. The
test are temporarily skipped since they detected quite a few bugs that
will be corrected with OTP-9792.
The following bug is corrected in this commit:'
reltool_server did not recognize {App,InclApps} inside a 'rel'
specification in the reltool config, e.g.
{rel, "myrel", "1.0", [{myapp,[app2]}]}.
|
|
OTP-9794
This commit adds a test suite for manual testing of the reltool
GUI. By use of test_server:break/1 function, it guides the user
through a set of tests in the GUI.
Some minor checks are also added to the automatic GUI test,
reltool_wx_SUITE.
|
|
OTP-9794
Test cases are added for the following functions:
* reltool_server:get_sys
* reltool_server:set_sys
* reltool_server:get_status
* reltool_server:gen_rel_files
* reltool:eval_target_spec
* reltool:get_server
|
|
OTP-9794
This test case revealed a bug that occurs when calling
reltool_server:get_mod after reltool_server:undo_config. get_mod reads
from the module table (ets) and not from the reltool_server state,
while undo_config only changes the state. This bug has been corrected,
so undo_config now updates both state and tables (it does the same as
set_sys).
|
|
OTP-9794
Some test cases in reltool_server_SUITE tried to create a target
system and then start a node running this system. If creation of the
target system failed, an attempt would still be made to start the
node. This has been corrected so the test is immediately terminated if
the target system could not be created.
|
|
OTP-9794
This is a test of the mechanism with which reltool_server decides
which applications to include and not, based on function calls between
applications. This is also the mechanism which forms the base for the
application- and module dependency graphs.
|
|
OTP-9794
The following test cases are added for the inteface from GUI to
reltool_server:
* get_config
* get_apps
* set_app_and_undo
* set_apps_and_undo
* load_config_and_undo
* reset_config_and_undo
* save_config
The following bugs were found and corrected:
* If set_apps failed, then the state of reltool_server would not
be reset to how it was before the failing operation - and every
operation done afterwards would also (seem to) fail.
* undo_config did not work after reset_config - since faulty #sys
record was stored as old_sys.
* undo_config did not work after set_app (used when changing the
content of an application from the GUI) - since old_sys was not
set. Also old_status was not set causing possible warnings to
disappear.
* undo_config did not work after set_apps (used e.g. when
excluding or including an application from the GUI) - since
old_sys was not set. Also old_status was not set causing
possible warnings to disappear.
|
|
When a release is generated and the applications in the release do not
define a value for the start_phases entry of their .app files, reltool
will generate the following entry in the .app files of the release:
{start_phases, undefined}
If this happens, when trying to create a release upgrade systools will
fail because it doesn't allow the start_phases entry to be set to
undefined. This patch avoids this situation by not generating a
start_phases entry when it is set to undefined.
|
|
|
|
* siri/reltool/doc-rootdir-recommendataion/OTP-9742:
Add recommendation about RootDir parameter to reltool:eval_target_spec/3
|
|
|
|
|
|
This reverts commit e21ff9b0b69219ab3853be7e80813156113152b7.
|
|
|
|
These dependency files was once used when building the documentation,
but are no longer needed.
|
|
|
|
* siri/reltool/empty-radiobox/OTP-9384:
Do not add an empty radio box on the releases tab for the start_clean release
|
|
First, the radiobox is changed to a listbox, since this will allow
multiple selections. This is however for future use - for now a
selection will only cause a printout in the erlang shell.
Second, add kernel and stdlib to the list of applications in order to
make the picture complete and avoid an empty list (radio) box for the
start_clean release.
|
|
If a module is duplicated in the library directories visible to
reltool, and the configuration does not point out which file to use,
then reltool:start will fail. This commit adds a pop-up which asks if
it should continue with a "safe" configuration:
[{incl_cond,exclude},
{app,kernel,[{incl_cond,include}]},
{app,stdlib,[{incl_cond,include}]},
{app,sasl,[{incl_cond,include}]}]
|
|
|
|
Earlier, reltool expected all module names detected under the lib
directories to have unique names. If this was not the case, the result
was undefined - i.e. the beam file of the duplicated module might be
included in multiple applications in the target area, or it might even
be excluded from all applications.
This commit adds awareness in reltool that a module might occur in
multiple applications, and it is allowed as long as the module or it's
application is explicitely excluded in all but one of the containing
applications.
|
|
* hw/call-chmod-without-f:
Call chmod without the "-f" flag
Conflicts:
erts/emulator/test/Makefile
lib/asn1/test/Makefile
lib/crypto/test/Makefile
lib/debugger/test/Makefile
lib/docbuilder/test/Makefile
lib/edoc/test/Makefile
lib/erl_interface/test/Makefile
lib/inviso/test/Makefile
lib/parsetools/test/Makefile
lib/percept/test/Makefile
lib/ssl/test/Makefile
lib/syntax_tools/test/Makefile
lib/test_server/test/Makefile
lib/tools/test/Makefile
OTP-9170
|
|
* siri/reltool/skip-xref-test-when-debug/OTP-9133:
Skip reltool_app_SUITE:undef_funcs on debug compiled emulator
|
|
This is the correction of the bug not allowing the values 'strip' or
'all' for the app_file option in reltool.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
While at it, applied some cleanups and code modernizations suggested by tidier.
|
|
|
|
|
|
The reltool module contained two seriously erroneous specs which caused
bogus warnings when dialyzing reltool and some correct code of users.
These were fixed (specs for start_link/1 and eval_server/3).
While at it, did some tidier cleanups and some cosmetic changes.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
"-f" is a non-standard chmod option which at least SGI IRIX and HP UX do
not support. As the only effect of the "-f" flag is to suppress warning
messages, it can be safely omitted.
|
