| Age | Commit message (Collapse) | Author |
|---|
|
This release bring major changes and enhancements to package dns and math/big.
=== Breaking changes
* math/big: refactoring AddFloat to accept variadic interface{}
Previously, the AddFloat function only accept two interface parameters.
This changes make the AddFloat to accept many interface parameters.
* math/big: refactor NewFloat to accept interface{} instead of float64 only
The following types is added and can be initialized to Float: []byte,
uint, uint16, uint32, uint64, big.Int, *big.Int, big.Rat, and *big.Rat.
* math/big: refactoring the Rat
The promise of this package is to provide the Rat type with RoundToZero
and without panic.
In order to do that we remove the MustRat() function and check for nil
on pointer receiver before doing operation, and check for zero value
when doing quotation.
Another breaking changes are,
* Humanize() with nil pointer receiver will return string "0"
* Any IsXxx() methods with nil pointer receiver will return false
* MarshalJSON() will nil pointer receiver will return "null" instead of
"0" now
* lib/dns: refactoring, change signature of client Lookup
Previously, Lookup() method on Client pass the question name, type, and
class as parameters.
This changes make those parameters into MessageQuestion.
* lib/dns: refactoring ZoneFile into Zone
Reason: A Zone is not always represented by file, it just that in this
package, it is.
This changes rename the type ZoneFile into Zone.
* lib/dns: refactoring message question
Previously the type for message question section SectionQuestion.
This changes, rename the type to MessageQuestion.
* lib/dns: refactoring the section header
This changes rename the SectionHeader into MessageHeader.
The pack() method is optimized with the following results,
benchmark old ns/op new ns/op delta
BenchmarkMessageHeader_pack-8 66.2 21.7 -67.31%
benchmark old allocs new allocs delta
BenchmarkMessageHeader_pack-8 3 1 -66.67%
benchmark old bytes new bytes delta
BenchmarkMessageHeader_pack-8 32 16 -50.00%
The unpack() method is simplified by minimizing the if-condition.
This changes also fix the pack and unpack OpCode for value other then 0,
due to wrong shift value, 2 instead of 3.
* lib/dns: refactoring, introduce new type RecordClass
Previously, the record class is represented by uint16 using prefix
QueryClassXxx.
This changes make the record class to be an independent type, to make
code more strict (prevent passing invalid value), and readable.
* lib/dns: refactoring, create type RecordType to represent type of RR
Previously, we use uint16 to represent type for ResourceRecord Type or
Question type.
To make the code more strict, where parameter or return value, must be
expected as record type, we add new type to represent the RR type:
RecordType.
This changes also rename any variable name of QType or qtype to RType
or rtype because QType is misleading. The type defined the
ResourceRecord to be queried not only question.
* lib/dns: remove the fallback name servers (NS) from server options
The original idea of fallback NS is to send the query to the one define
to original resolv.conf, instead of using the one defined by user in
ServerOptions NameServers, when an error occured.
But, most of error usually caused by network (disconnected, time out),
so re-sending query to fallback NS does not have any effect if the network
it self is not working.
This changes remove the unnecessary and complex fallback NS from
server.
=== New features
* lib/dns: add support to save and load caches to/from storage
The CachesSave method write the non-local answers into an io.Writer,
encoded with gob.
The CachesLoad method load the gob encoded answers from an io.Reader.
* lib/ssh: use agent defined by config or from environment variable
Previously, we only check the environment variable SSH_AUTH_SOCK
to decide whether the client identity will fetched from agent or
from private key file.
This changes use the GetIdentityAgent from config Section to derive the
path to unix socket. It will return an empty string if the IdentityAgent
is set to "none" in the Section or when SSH_AUTH_SOCK is not exist or
empty.
* ssh/config: add support for section variable IdentityAgent
There are four possible value for IdentityAgent: SSH_AUTH_SOCK,
<$STRING>, <PATH>, or "none".
If SSH_AUTH_SOCK, the socket path is read from the environment variable
SSH_AUTH_SOCK.
If value start with "$", then the socket path is set based on value of
that environment variable.
Other string beside "none" will be considered as path to socket.
=== Bug fixes
* math/big: return nil on Quo and QuoRat instead of panic
Previously, if the first parameter of Quo or the second/next parameters
of QuoRat is not convertable to Rat or zero, the method/function will
panic.
This changes make it less intrusive, instead of panic we check for
zero value and return nil immediately.
* lib/dns: do not cache truncated answer
Previously only answer with non-zero response code is ignored.
This changes ignore also answer where response header is truncated.
=== Enhancements
* lib/dns: realign all struct fields
Turns out the struct that we frequently used, answer and resource_record,
is not optimized.
answer.go:15:13: struct with 56 pointer bytes could be 24 = 32 bytes
...
resource_record.go:24:21: struct with 56 pointer bytes could be 40 = 16
This changes reorder all structs field to save space in memory.
* lib/dns: make the TCP forwarders as complementary of UDP
The TCP forwarders only active when client send the DNS request as TCP.
When the server receive that request it should also forward the request
as TCP not as UDP to prevent the truncated response.
Another use case for TCP is when the response is truncated, the client
will send the query back through TCP connection. The server should
forward this request using TCP instead of UDP.
* lib/dns: use different packet between UDP and TCP messages
Previously, all packet size for reading and sending the message is
fixed to 4096, even on UDP.
This changes set the UDP packet size maximum to 512 bytes and others to
4096 bytes.
While at it, minimize copying packet if its not reusable inside a method.
|
|
Previously, the AddFloat function only accept two interface parameters.
This changes make the AddFloat to accept many interface parameters.
|
|
The following types is added and can be initialized to Float: []byte,
uint, uint16, uint32, uint64, big.Int, *big.Int, big.Rat, and *big.Rat.
|
|
The following alignment reported and fixed by fieldaligment tool,
float_test.go:103:13: struct with 16 pointer bytes could be 8
float_test.go:254:13: struct with 16 pointer bytes could be 8
big/float_test.go:302:13: struct with 32 pointer bytes could be 16
|
|
The promise of this package is to provide the Rat type with RoundToZero
and without panic.
In order to do that we remove the MustRat() function and check for nil
on pointer receiver before doing operation, and check for zero value
when doing quotation.
Another breaking changes are,
* Humanize() with nil pointer receiver will return string "0"
* Any IsXxx() methods with nil pointer receiver will return false
* MarshalJSON() will nil pointer receiver will return "null" instead of
"0" now
|
|
Previously, if the first parameter of Quo or the second/next parameters
of QuoRat is not convertable to Rat or zero, the method/function will
panic.
This changes make it less intrunsive, instead of panic we check for
zero value and return nil immediately.
|
|
Previously, Lookup() method on Client pass the question name, type, and
class as parameters.
This changes make those parameters into MessageQuestion.
|
|
|
|
|
|
Reason: A Zone is not always represented by file, it just that in this
package, it is.
This changes rename the type ZoneFile into Zone.
|
|
|
|
Using the fieldalignment, we got the following output,
answer.go:15:13: struct with 56 pointer bytes could be 24 = 32 bytes
caches.go:27:13: struct with 24 pointer bytes could be 16 = 8
dot_client.go:20:16: struct with 16 pointer bytes could be 8 = 8
hosts_file.go:24:16: struct with 64 pointer bytes could be 48 = 16
message.go:53:14: struct of size 176 could be 168 = 8
rdata_mx.go:20:14: struct with 16 pointer bytes could be 8 = 8
rdata_opt.go:18:15: struct with 16 pointer bytes could be 8 = 8
rdata_srv.go:16:15: struct with 64 pointer bytes could be 56 = 8
rdata_wks.go:30:15: struct with 40 pointer bytes could be 32 = 8
request.go:20:14: struct with 32 pointer bytes could be 24 = 8
resource_record.go:24:21: struct with 56 pointer bytes could be 40 = 16
server.go:76:13: struct with 120 pointer bytes could be 112 = 8
server_options.go:20:20: struct of size 240 could be 224 = 16
tcp_client.go:21:16: struct with 40 pointer bytes could be 24 = 16
udp_client.go:24:16: struct with 24 pointer bytes could be 16 = 8
udp_client_pool.go:25:20: struct with 24 pointer bytes could be 16 = 8
zone_file.go:22:15: struct with 128 pointer bytes could be 120 = 8
zone_parser.go:53:17: struct with 88 pointer bytes could be 72 = 16
Turns out the struct that we frequently used, answer and resource_record,
is not optimized.
This changes reorder all structs field to save space in memory.
|
|
Previously the type for message question section SectionQuestion.
This changes, rename the type to MessageQuestion.
|
|
This changes rename the SectionHeader into MessageHeader.
The pack() method is optimized with the following results,
benchmark old ns/op new ns/op delta
BenchmarkMessageHeader_pack-8 66.2 21.7 -67.31%
benchmark old allocs new allocs delta
BenchmarkMessageHeader_pack-8 3 1 -66.67%
benchmark old bytes new bytes delta
BenchmarkMessageHeader_pack-8 32 16 -50.00%
The unpack() method is simplified by minimizing the if-condition.
This changes also fix the pack and unpack OpCode for value other then 0,
due to wrong shift value, 2 instead of 3.
|
|
Previously, the record class is represented by uint16 using prefix
QueryClassXxx.
This changes make the record class to be an independent type, to make
code more strict (prevent passing invalid value), and readable.
|
|
Previously, we use uint16 to represent type for ResourceRecord Type or
Question type.
To make the code more strict, where parameter or return value, must be
expected as record type, we add new type to represent the RR type:
RecordType.
This changes also rename any variable name of QType or qtype to RType
or rtype because QType is misleading. The type defined the
ResourceRecord to be queried not only question.
|
|
The ini method initialize the SOA record by setting fields to its default
value if its empty.
The default value for RName is "root", for Serial is current epoch,
for Refresh is 1 day, for Retry is 1 hour, and for Minimum is 1 hour.
While at it, update the comments to be more understandable and represent
the type or fields.
|
|
The off field previously used to record the next index after parsing
domain name with offset. Since the field only used once, after calling
unpackDomainName, we return the off value instead as "end".
This changes make the unpackDomainName() become a function.
This changes rename field offTTL to idxTTL to match with their value.
This changes alos pefix all errors with the method or function names.
|
|
|
|
|
|
The TCP forwarders only active when client send the DNS request as TCP.
When the server receive that request it should also forward the request
as TCP not as UDP to prevent the truncated response.
Another use case for TCP is when the response is truncated, the client
will send the query back through TCP connection. The server should
forward this request using TCP instead of UDP.
|
|
The original idea of fallback NS is to send the query to the one define
to original resolv.conf, instead of using the one defined by user in
ServerOptions NameServers, when an error occured.
But, most of error usually caused by network (disconnected, time out),
so re-sending query to fallback NS does not have any effect if the network
it self is not working.
This changes remove the unnecessary and complex fallback NS from
server.
|
|
Using snake case make the file naming more readable.
|
|
|
|
Instead of returning four variables, the parseNameServers now have
parameter isPrimary to set the primary address or fallback address.
|
|
Previously only answer with non-zero response code is ignored.
This changes ignore also answer where response header is truncated.
|
|
Previously, all packet size for reading and sending the message is
fixed to 4096, even on UDP.
This changes set the UDP packet size maximum to 512 bytes and others to
4096 bytes.
While at it, minimize copying packet if its not reusable inside a method.
|
|
The CachesSave method write the non-local answers into an io.Writer,
encoded with gob.
The CachesLoad method load the gob encoded answers from an io.Reader.
|
|
Previously, we only check the environment variable SSH_AUTH_SOCK
to decide whether the client identity will fetched from agent or
from private key file.
This changes use the GetIdentityAgent from config Section to derive the
path to unix socket. It will return an empty string if the IdentityAgent
is set to "none" in the Section or when SSH_AUTH_SOCK is not exist or
empty.
|
|
There are four possible value for IdentityAgent: SSH_AUTH_SOCK,
<$STRING>, <PATH>, or "none".
If SSH_AUTH_SOCK, the socket path is read from the environment variable
SSH_AUTH_SOCK.
If value start with "$", then the socket path is set based on value of
that environment variable.
Other string beside "none" will be considered as path to socket.
|
|
=== Breaking changes
* lib/memfs: move the embedded parameter to Options
Since the GoEmbed can be called only when MemFS instance is initiated,
it would be better if parameters for GoEmbed also initialized in the
struct Options.
In this way any additional parameters needed to add to GoEmbed does not
changes the method signature in the future.
This commit add new type EmbedOptions that contains the parameters
for GoEmbed.
In this new type, we add new field EmbedWithoutModTime.
if its true, the modification time for all files and directories are not
stored inside generated code, instead all files will use the current
time when the program is running.
* lib/totp: make the New to accept only hash with SHA1, SHA256, or SHA512
Previously, the first parameter to New is a function that return
hash.Hash. This signature can be misleading, because md5.New also
can return hash.Hash but not usable in TOTP.
This changes update the New function signature to accept defined
type that can be set to SHA1, SHA256, or SHA512.
* lib/bytes: refactoring and cleaning up the bytes package
The bytes package, and many other packages in this module, is written
when I still learning and using the Go language for master thesis.
Some of the code, like function signature, does not follow the
Go idiom, at least not how the Go source code looks like.
A breaking changes,
* WriteUint16 and WriteUint32 accept slice only without pointer.
There is no need to pass slice as pointer to function if we want
to modify the content, as long as the backing storage is not
changed.
Bug fixes,
* PrintHex: fix print layout on the last line
* ReadHexByte: fix possible index out of range
* SkipAfterToken return -1 and false if no token found, as promised
in the comment, instead of the length of text.
We move all unit test to example so we have test and example in the
documentation at the same time.
This changes make all test coverage 100%.
* lib/bytes: refactoring AppendXxx functions
Previously, we pass pointer to slice on AppendInt16, AppendInt32,
AppendUint16, and AppendUint32 functions. This model of function
signature is not a Go idiom. It is written when I am still new to
Go.
* lib/ascii: change signature of ToLower and ToUpper
Using pointer to slice on method or function is not a Go idiom.
It is created when I still new to Go.
* lib/memfs: refactoring Node field V into Content
The reason why the field named V is because it's short. It's come
from my C/C++ experience that got carried away when writing this
package.
Now, after having more time writing Go, I prefer clarity over
cleverity(?).
* lib/memfs: set the node modification time in embedded file
This changes set all node modification time in embedded files to
the node modTime using Unix() and Nanosecond() values.
Since the time will always changes we need to remove the test to
generate file gen_test.go to prevent the file being modified and
re-adding the same file every time we run local tests.
=== New features
* lib/ini: add function IsValidVarName
The IsValidVarName check if "v" is valid variable name, where the
first character must be a letter and the rest should contains only
letter, digit, period, hyphen, or underscore.
If "v" is valid it will return true.
* lib/memfs: set the node modification time in embedded file
This changes set all node modification time in embedded files to
the node modTime using Unix() and Nanosecond() values.
Since the time will always changes we need to remove the test to
generate file gen_test.go to prevent the file being modified and
re-adding the same file every time we run local tests.
* lib/io: add method String to FileState type
The String method return the string representation of FileState.
Usually used only for debugging.
* lib/smtp: implement method Noop on Client
Noop send the NOOP command to server with optional message.
On success, it will return response with Code 250, StatusOK.
While at it fix double call to recv on Reset() method.
* lib/smtp: implement method Reset on Client
The Reset() method send the STMP RSET command to the server.
This command clear the current buffer on MAIL, RCPT, and DATA, but not the
EHLO/HELO buffer.
On success, it will return response with Code 250, StatusOK.
=== Bug fixes
* lib/ascii: fix IsHex return false on 0
* lib/memfs: fix parent empty directory not removed
Use case:
x
x/y
If x/y is empty, and x processed first, the x will
not be removed.
This commit fix this, by sorting the paths in descending order first
to make empty parent removed clearly. In above case the order
or check become,
x/y
x
While at it, update an example of New to give preview of input and
what the expected output for certain operations.
* lib/xmlrpc: rewrite the Client connection using lib/http
Using socket connection require reading the HTTP response header before
we can process the response body.
Instead of rewrite the parser, use the lib/http to send and receive
the request/response.
* lib/io: do not use absolute path on dummy Watcher parent SysPath
Converting that parameter path to absolute path may cause unpredictable
result on module that use it.
=== Chores
* lib/ini: add example of marshal/unmarshaling of section with sub
This changes also fix the example of field tag on marshaling the
map.
* lib/io: add method String to FileState type
The String method return the string representation of FileState.
Usually used only for debugging.
* lib/memfs: remove unnecessary initialization on NewNode
The zero value for V ([]byte) is already nil and Node.Childs
([]*Node) does not need to be initialized with make if size is 0.
* lib/io: use t.Cleanup instead of defer on test
Signed-off-by: Shulhan <ms@kilabit.info>
|
|
This is to remove "embed_test" being displayed on the documentation page.
|
|
Since the GoEmbed can be called only when MemFS instance is initiated,
it would be better if parameters for GoEmbed also initialized in the
struct Options.
In this way any additional parameters needed to add to GoEmbed does not
changes the method signature in the future.
This commit add new type EmbedOptions that contains the parameters
for GoEmbed.
In this new type, we add new field EmbedWithoutModTime.
if its true, the modification time for all files and directories are not
stored inside generated code, instead all files will use the current
time when the program is running.
|
|
Previously, the first parameter to New is a function that return
hash.Hash. This signature can be misleading, because md5.New also
can return hash.Hash but not usable in TOTP.
This changes update the New function signature to accept defined
type that can be set to SHA1, SHA256, or SHA512.
|
|
|
|
The IsValidVarName check if "v" is valid variable name, where the
first character must be a letter and the rest should contains only
letter, digit, period, hyphen, or underscore.
If "v" is valid it will return true.
|
|
This changes also fix the example of field tag on marshaling the
map.
|
|
The bytes package, and many other packages in this module, is written
when I still learning and using the Go language for master thesis.
Some of the code, like function signature, does not follow the
Go idiom, at least not how the Go source code looks like.
A breaking changes,
* WriteUint16 and WriteUint32 accept slice only without pointer.
There is no need to pass slice as pointer to function if we want
to modify the content, as long as the backing storage is not
changed.
Bug fixes,
* PrintHex: fix print layout on the last line
* ReadHexByte: fix possible index out of range
* SkipAfterToken return -1 and false if no token found, as promised
in the comment, instead of the length of text.
We move all unit test to example so we have test and example in the
documentation at the same time.
This changes make all test coverage 100%.
|
|
Previously, we pass pointer to slice on AppendInt16, AppendInt32,
AppendUint16, and AppendUint32 functions. This model of function
signature is not a Go idiom. It is written when I am still new to
Go.
|
|
Using pointer to slice on method or function is not a Go idiom.
It is created when I still new to Go.
|
|
This is including IsAlnum, IsAlpha, IsDigit, IsDigits, IsHex, IsSpace,
and Random.
|
|
|
|
The variable memFS is exist only on test file, but if the file name
for generated GoEmbed is not _test.go the "go install /lib/memfs/..."
will thrown an error
go install ./...
# github.com/shuLhan/share/lib/memfs/embed_test
lib/memfs/embed_test/memfs_embed.go:20:31: undefined: memFS
lib/memfs/embed_test/memfs_embed.go:21:31: undefined: memFSR
|
|
On the test we re-create the root directory with sub-directory.
This action will trigger two FileStateCreated events, one for root "/"
and another one for sub-directory "/assets".
|
|
The generated comment for _getNode should match with generated
function name which has the prefix of variable name.
|
|
Previously, the PathNode has two fields to store the node in memory,
one is "v" that store map of path to *Node and another is "f" that
store the map of path to function that return a *Node.
The "f" is used to handle embedding Go generated code, but since
the template now can handle generated dynamic path and Node this field
is not used anymore.
|
|
One of the hardest when multiple packages is used is to detect where
there error happened. For example, if a program return an error io.EOF,
it's hard to detect the exact method or function that caused its,
especially when processing multiple files with network connection.
|
|
This changes rename the receiver name on Node from "leaf" to "node".
|
|
Previously we use the "testdata" for output of Node.Save() test.
This is caused the modified time on "testdata" get changes every
time we run the test and cause the embed_test failed.
This changes use temporary directory and revert part of Node Save()
unit test from commit 792ba17ac1f.
|
|
The reason why the field named V is because it's short. It's come
from my C/C++ experience that got carried away when writing this
package.
Now, after having more time writing Go, I prefer clarity over
cleverity(?).
|
