[PATCH 4/4] imap-send: use curl by default

Signed-off-by: Nicolas Morey-Chaisemartin <redacted>
---
 imap-send.c | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/imap-send.c b/imap-send.c
index 90b8683ed..4ebc16437 100644
--- a/imap-send.c
+++ b/imap-send.c

@@ -35,13 +35,7 @@ typedef void *SSL;
 #include "http.h"
 #endif
 
-#if defined(USE_CURL_FOR_IMAP_SEND) && defined(NO_OPENSSL)
-/* only available option */
 #define USE_CURL_DEFAULT 1
-#else
-/* strictly opt in */
-#define USE_CURL_DEFAULT 0
-#endif
 
 static int verbosity;
 static int use_curl = USE_CURL_DEFAULT;

-- 
2.14.0.rc1.16.g87fcec1e8

On 7 August 2017 at 16:04, Nicolas Morey-Chaisemartin
[off-list ref] wrote:

So this is now basically "static int use_curl = 1;".

Do we need a compile-time escape-hatch in case someone really needs
to avoid curl, e.g., because they have a too old version? I suppose
there is a conceptual difference between the "default", i.e., the value
of USE_CURL_DEFAULT that is assigned to "use_curl", and the "default
default", i.e., the value that is normally assigned to USE_CURL_DEFAULT.

Martin

Le 07/08/2017 à 18:37, Martin Ågren a écrit :

On 7 August 2017 at 16:04, Nicolas Morey-Chaisemartin
[off-list ref] wrote:

quoted

Signed-off-by: Nicolas Morey-Chaisemartin <redacted>
---
 imap-send.c | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/imap-send.c b/imap-send.c
index 90b8683ed..4ebc16437 100644
--- a/imap-send.c
+++ b/imap-send.c

@@ -35,13 +35,7 @@ typedef void *SSL;
 #include "http.h"
 #endif

-#if defined(USE_CURL_FOR_IMAP_SEND) && defined(NO_OPENSSL)
-/* only available option */
 #define USE_CURL_DEFAULT 1
-#else
-/* strictly opt in */
-#define USE_CURL_DEFAULT 0
-#endif

 static int verbosity;
 static int use_curl = USE_CURL_DEFAULT;

So this is now basically "static int use_curl = 1;".

Do we need a compile-time escape-hatch in case someone really needs
to avoid curl, e.g., because they have a too old version? I suppose
there is a conceptual difference between the "default", i.e., the value
of USE_CURL_DEFAULT that is assigned to "use_curl", and the "default
default", i.e., the value that is normally assigned to USE_CURL_DEFAULT.

Martin

The curl code depends on USE_CURL_FOR_IMAP_SEND so even with use_curl == 1, it won't be an issue for people without curl (or old one).
I wasn't sure whether to drop the define or not and figure it might be worth keeping in case in change in the future for some reason.
I don't mind dropping it and hardcofing the default to 1

Also on a side note, I have a case where authentication works with --no-curl but fails withs --curl. Still trying to figure out why.

Nicolas

On 7 August 2017 at 19:10, Nicolas Morey-Chaisemartin
[off-list ref] wrote:

Le 07/08/2017 à 18:37, Martin Ågren a écrit :

quoted

On 7 August 2017 at 16:04, Nicolas Morey-Chaisemartin
[off-list ref] wrote:

quoted

Signed-off-by: Nicolas Morey-Chaisemartin <redacted>
---
 imap-send.c | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/imap-send.c b/imap-send.c
index 90b8683ed..4ebc16437 100644
--- a/imap-send.c
+++ b/imap-send.c

@@ -35,13 +35,7 @@ typedef void *SSL;
 #include "http.h"
 #endif

-#if defined(USE_CURL_FOR_IMAP_SEND) && defined(NO_OPENSSL)
-/* only available option */
 #define USE_CURL_DEFAULT 1
-#else
-/* strictly opt in */
-#define USE_CURL_DEFAULT 0
-#endif

 static int verbosity;
 static int use_curl = USE_CURL_DEFAULT;

So this is now basically "static int use_curl = 1;".

Do we need a compile-time escape-hatch in case someone really needs
to avoid curl, e.g., because they have a too old version? I suppose
there is a conceptual difference between the "default", i.e., the value
of USE_CURL_DEFAULT that is assigned to "use_curl", and the "default
default", i.e., the value that is normally assigned to USE_CURL_DEFAULT.

Martin

The curl code depends on USE_CURL_FOR_IMAP_SEND so even with use_curl == 1, it won't be an issue for people without curl (or old one).

I have just looked at the sources and haven't thought too hard about it,
but doesn't it mean that compiling without USE_CURL_FOR_IMAP_SEND
results in a binary such that you must use --no-curl or get used to seeing
"warning: --curl not supported in this build"?

I wasn't sure whether to drop the define or not and figure it might be worth keeping in case in change in the future for some reason.
I don't mind dropping it and hardcofing the default to 1

I did not intend to suggest that. Just to be clear, I am very unfamiliar
with most of the Git codebase. Please don't take anything I say as
advice. :) As a question about something you have or haven't already
thought about, sure. :)

Martin

On Mon, Aug 07, 2017 at 04:04:05PM +0200, Nicolas Morey-Chaisemartin wrote:

Signed-off-by: Nicolas Morey-Chaisemartin <redacted>

Thanks for moving forward with this.

Can you please flesh out your commit messages with some of the reasoning
and related discussion? I know from a nearby thread why we want to flip
the default, but people reading `git log` much later will not have that
context.

I agree with the comments Martin made here. I think there are really two
levels of "default" we need to care about:

  1. Build-time: do we default to requiring curl if you want imap-send
     at all?

  2. Run-time: if build with both implementations, which do we use by
     default? Related, if there is only one implementation, what should
     the default be?

I think the answer to (1) is that we still want to build imap-send
without USE_CURL_FOR_IMAP_SEND if the user doesn't have curl installed.
And your patch leaves that, which is good. Though if we are deprecating
it, we may want to issue a deprecation warning (eventually; we can still
switch the run-time default now, get more data on whether a
deprecation/switch is a good idea, and then later decide to deprecate).

For (2), you're trying to switch the default when both are built. But I
think it's important to continue to default to the old-style
implementation if that's the only thing that was built. Otherwise it
effectively becomes the build-time deprecation warning, and we're not
quite ready for that.

So I think this maybe needs to be:

  #if defined(USE_CURL_FOR_IMAP_SEND)
  /* Always default to curl if it's available. */
  #define USE_CURL_DEFAULT 1
  #else
  /* We don't have curl, so continue to use the historical implementation */
  #define USE_CURL_DEFAULT 0
  #endif

-Peff