* [Tarantool-patches] [PATCH 1/1] app: os.setenv() affects os.environ()
@ 2020-02-06 22:47 Vladislav Shpilevoy
2020-02-10 11:03 ` Igor Munkin
2020-02-19 21:42 ` Nikita Pettik
0 siblings, 2 replies; 4+ messages in thread
From: Vladislav Shpilevoy @ 2020-02-06 22:47 UTC (permalink / raw)
To: tarantool-patches, imun, korablev
os.setenv() and os.environ() are Lua API for
extern char **environ;
int setenv();
The Open Group standardized access points for environment
variables. But there is no a word about that environ never
changes. Programs can't relay on that. For example, addition of
a new variable may cause realloc of the whole environ array, and
therefore change of its pointer value. That was exactly the case
in os.environ() - it was using value of environ array remembered
when Tarantool started.
And os.setenv() could realloc the array and turn the saved pointer
into garbage.
Closes #4733
---
Issue: https://github.com/tarantool/tarantool/issues/4733
Branch: https://github.com/tarantool/tarantool/tree/gerold103/gh-4733-os.environ
src/lua/env.lua | 3 +--
test/app/env.result | 46 +++++++++++++++++++++++++++++++++++++++----
test/app/env.test.lua | 27 +++++++++++++++++++++----
3 files changed, 66 insertions(+), 10 deletions(-)
diff --git a/src/lua/env.lua b/src/lua/env.lua
index b150a48a9..dd1616a84 100644
--- a/src/lua/env.lua
+++ b/src/lua/env.lua
@@ -9,9 +9,8 @@ ffi.cdef[[
int unsetenv(const char *name);
]]
-local environ = ffi.C.environ
-
os.environ = function()
+ local environ = ffi.C.environ
if not environ then
return nil
end
diff --git a/test/app/env.result b/test/app/env.result
index f725b01c0..c584947fb 100644
--- a/test/app/env.result
+++ b/test/app/env.result
@@ -24,6 +24,9 @@ type(env_dict)
---
- table
...
+err = nil
+---
+...
test_run:cmd("setopt delimiter ';'")
---
- true
@@ -31,11 +34,46 @@ test_run:cmd("setopt delimiter ';'")
do
for k, v in pairs(env_dict) do
if type(k) ~= 'string' or type(v) ~= 'string' then
- return false
+ err = {k, v}
+ break
end
end
- return true
-end;
+end
+test_run:cmd("setopt delimiter ''");
---
-- true
+...
+err
+---
+- null
+...
+--
+-- gh-4733: os.setenv() should affect os.environ
+--
+size = 0
+---
+...
+for _, __ in pairs(os.environ()) do size = size + 1 end
+---
+...
+for i = 1, size do os.setenv('gh-4733-test-var'..i, tostring(i)) end
+---
+...
+env = os.environ()
+---
+...
+err = nil
+---
+...
+for i = 1, size do \
+ local value = env['gh-4733-test-var'..i] \
+ if value ~= tostring(i) then \
+ err = {i, value} \
+ break \
+ end \
+end
+---
+...
+err
+---
+- null
...
diff --git a/test/app/env.test.lua b/test/app/env.test.lua
index fbc1ace86..b725f0124 100644
--- a/test/app/env.test.lua
+++ b/test/app/env.test.lua
@@ -9,13 +9,32 @@ do os.getenv('location') end
env_dict = os.environ()
type(env_dict)
+err = nil
test_run:cmd("setopt delimiter ';'")
do
for k, v in pairs(env_dict) do
if type(k) ~= 'string' or type(v) ~= 'string' then
- return false
+ err = {k, v}
+ break
end
end
- return true
-end;
-test_run:cmd("setopt delimiter ''")
+end
+test_run:cmd("setopt delimiter ''");
+err
+
+--
+-- gh-4733: os.setenv() should affect os.environ
+--
+size = 0
+for _, __ in pairs(os.environ()) do size = size + 1 end
+for i = 1, size do os.setenv('gh-4733-test-var'..i, tostring(i)) end
+env = os.environ()
+err = nil
+for i = 1, size do \
+ local value = env['gh-4733-test-var'..i] \
+ if value ~= tostring(i) then \
+ err = {i, value} \
+ break \
+ end \
+end
+err
--
2.21.1 (Apple Git-122.3)
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Tarantool-patches] [PATCH 1/1] app: os.setenv() affects os.environ()
2020-02-06 22:47 [Tarantool-patches] [PATCH 1/1] app: os.setenv() affects os.environ() Vladislav Shpilevoy
@ 2020-02-10 11:03 ` Igor Munkin
2020-02-19 21:42 ` Nikita Pettik
1 sibling, 0 replies; 4+ messages in thread
From: Igor Munkin @ 2020-02-10 11:03 UTC (permalink / raw)
To: Vladislav Shpilevoy; +Cc: tarantool-patches
Vlad,
Thanks for the patch, LGTM.
Minor: I see no particular reason to implement the test below in "diff"
instead of TAP, but I don't insist since all other related tests are
already implemented in "diff".
Side note: OMG, I guess spoiling Lua standart namespaces is some kind of
obsession. I can understand the reason of adding new methods in string
table since it's a metatable for all string objects. However, I can't
understand this activity regarding other standart namespaces starting
from Mike with his table.new (but thanks for an additional 'require'
required to enabled this feature) and finishing with #1718. I hope
sometime we move everything to a separate namespace like the one uJIT
has. The remark is not related to the patch, the changes are good,
thanks again.
On 06.02.20, Vladislav Shpilevoy wrote:
> os.setenv() and os.environ() are Lua API for
>
> extern char **environ;
> int setenv();
>
> The Open Group standardized access points for environment
> variables. But there is no a word about that environ never
> changes. Programs can't relay on that. For example, addition of
> a new variable may cause realloc of the whole environ array, and
> therefore change of its pointer value. That was exactly the case
> in os.environ() - it was using value of environ array remembered
> when Tarantool started.
>
> And os.setenv() could realloc the array and turn the saved pointer
> into garbage.
>
> Closes #4733
> ---
> Issue: https://github.com/tarantool/tarantool/issues/4733
> Branch: https://github.com/tarantool/tarantool/tree/gerold103/gh-4733-os.environ
>
> src/lua/env.lua | 3 +--
> test/app/env.result | 46 +++++++++++++++++++++++++++++++++++++++----
> test/app/env.test.lua | 27 +++++++++++++++++++++----
> 3 files changed, 66 insertions(+), 10 deletions(-)
>
> diff --git a/src/lua/env.lua b/src/lua/env.lua
> index b150a48a9..dd1616a84 100644
> --- a/src/lua/env.lua
> +++ b/src/lua/env.lua
> @@ -9,9 +9,8 @@ ffi.cdef[[
> int unsetenv(const char *name);
> ]]
>
> -local environ = ffi.C.environ
> -
> os.environ = function()
> + local environ = ffi.C.environ
> if not environ then
> return nil
> end
> diff --git a/test/app/env.result b/test/app/env.result
> index f725b01c0..c584947fb 100644
> --- a/test/app/env.result
> +++ b/test/app/env.result
> @@ -24,6 +24,9 @@ type(env_dict)
> ---
> - table
> ...
> +err = nil
> +---
> +...
> test_run:cmd("setopt delimiter ';'")
> ---
> - true
> @@ -31,11 +34,46 @@ test_run:cmd("setopt delimiter ';'")
> do
> for k, v in pairs(env_dict) do
> if type(k) ~= 'string' or type(v) ~= 'string' then
> - return false
> + err = {k, v}
> + break
> end
> end
> - return true
> -end;
> +end
> +test_run:cmd("setopt delimiter ''");
> ---
> -- true
> +...
> +err
> +---
> +- null
> +...
> +--
> +-- gh-4733: os.setenv() should affect os.environ
> +--
> +size = 0
> +---
> +...
> +for _, __ in pairs(os.environ()) do size = size + 1 end
> +---
> +...
> +for i = 1, size do os.setenv('gh-4733-test-var'..i, tostring(i)) end
> +---
> +...
> +env = os.environ()
> +---
> +...
> +err = nil
> +---
> +...
> +for i = 1, size do \
> + local value = env['gh-4733-test-var'..i] \
> + if value ~= tostring(i) then \
> + err = {i, value} \
> + break \
> + end \
> +end
> +---
> +...
> +err
> +---
> +- null
> ...
> diff --git a/test/app/env.test.lua b/test/app/env.test.lua
> index fbc1ace86..b725f0124 100644
> --- a/test/app/env.test.lua
> +++ b/test/app/env.test.lua
> @@ -9,13 +9,32 @@ do os.getenv('location') end
>
> env_dict = os.environ()
> type(env_dict)
> +err = nil
> test_run:cmd("setopt delimiter ';'")
> do
> for k, v in pairs(env_dict) do
> if type(k) ~= 'string' or type(v) ~= 'string' then
> - return false
> + err = {k, v}
> + break
> end
> end
> - return true
> -end;
> -test_run:cmd("setopt delimiter ''")
> +end
> +test_run:cmd("setopt delimiter ''");
> +err
> +
> +--
> +-- gh-4733: os.setenv() should affect os.environ
> +--
> +size = 0
> +for _, __ in pairs(os.environ()) do size = size + 1 end
> +for i = 1, size do os.setenv('gh-4733-test-var'..i, tostring(i)) end
> +env = os.environ()
> +err = nil
> +for i = 1, size do \
> + local value = env['gh-4733-test-var'..i] \
> + if value ~= tostring(i) then \
> + err = {i, value} \
> + break \
> + end \
> +end
> +err
> --
> 2.21.1 (Apple Git-122.3)
>
--
Best regards,
IM
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Tarantool-patches] [PATCH 1/1] app: os.setenv() affects os.environ()
2020-02-06 22:47 [Tarantool-patches] [PATCH 1/1] app: os.setenv() affects os.environ() Vladislav Shpilevoy
2020-02-10 11:03 ` Igor Munkin
@ 2020-02-19 21:42 ` Nikita Pettik
2020-02-19 22:44 ` Vladislav Shpilevoy
1 sibling, 1 reply; 4+ messages in thread
From: Nikita Pettik @ 2020-02-19 21:42 UTC (permalink / raw)
To: Vladislav Shpilevoy; +Cc: tarantool-patches
On 06 Feb 23:47, Vladislav Shpilevoy wrote:
> os.setenv() and os.environ() are Lua API for
>
> extern char **environ;
> int setenv();
>
> The Open Group standardized access points for environment
> variables. But there is no a word about that environ never
> changes. Programs can't relay on that. For example, addition of
> a new variable may cause realloc of the whole environ array, and
> therefore change of its pointer value. That was exactly the case
> in os.environ() - it was using value of environ array remembered
> when Tarantool started.
>
> And os.setenv() could realloc the array and turn the saved pointer
> into garbage.
>
> Closes #4733
> ---
LGTM
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Tarantool-patches] [PATCH 1/1] app: os.setenv() affects os.environ()
2020-02-19 21:42 ` Nikita Pettik
@ 2020-02-19 22:44 ` Vladislav Shpilevoy
0 siblings, 0 replies; 4+ messages in thread
From: Vladislav Shpilevoy @ 2020-02-19 22:44 UTC (permalink / raw)
To: Nikita Pettik; +Cc: tarantool-patches
Hi! Thanks all for the reviews!
Pushed to master, 2.3, 2.2 and 1.10.
Also added a changelog record to the releases draft.
And here too, since I forgot to add it to the cover
letter (or that rule was added after my patch, don't
remember).
@ChangeLog
- os.environ() was not changed when os.setenv() was
called (gh-4733).
On 19/02/2020 22:42, Nikita Pettik wrote:
> On 06 Feb 23:47, Vladislav Shpilevoy wrote:
>> os.setenv() and os.environ() are Lua API for
>>
>> extern char **environ;
>> int setenv();
>>
>> The Open Group standardized access points for environment
>> variables. But there is no a word about that environ never
>> changes. Programs can't relay on that. For example, addition of
>> a new variable may cause realloc of the whole environ array, and
>> therefore change of its pointer value. That was exactly the case
>> in os.environ() - it was using value of environ array remembered
>> when Tarantool started.
>>
>> And os.setenv() could realloc the array and turn the saved pointer
>> into garbage.
>>
>> Closes #4733
>> ---
>
> LGTM
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2020-02-19 22:44 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-02-06 22:47 [Tarantool-patches] [PATCH 1/1] app: os.setenv() affects os.environ() Vladislav Shpilevoy
2020-02-10 11:03 ` Igor Munkin
2020-02-19 21:42 ` Nikita Pettik
2020-02-19 22:44 ` Vladislav Shpilevoy
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox