Skip to content

API Behavior Changes in v6.0.6 ​

This page lists every deliberate change in v6.0.6 that an existing integration can observe, so you can check your code against it before upgrading.

Tier 1: changes that can break an integration ​

Subscribers ​

subscriber.subscribe no longer re-subscribes an unsubscribed contact ​

This is a compliance fix. Before v6.0.6, subscriber.subscribe always re-subscribed an email address that had unsubscribed from the list: the status went back to Subscribed (or Opt-In Pending on a double opt-in list), the list suppression entry was deleted, and the response was identical to a new subscription. Any integration that synced contacts through this command silently undid opt-outs, which exposes the sender under CAN-SPAM and GDPR (octeth/oempro#3881).

The call is now refused with the new ErrorCode 12 and "PreviousSubscriptionStatus": "Unsubscribed". The subscriber row, its subscription source and its list suppression entry are left as they were, and no confirmation email is sent. A new request parameter, UpdateIfUnsubscribed, restores the old behaviour for one call when set to true.

Every success response also carries the new PreviousSubscriptionStatus field: empty for a new contact, Unsubscribed when UpdateIfUnsubscribed=true re-subscribed it.

Not affected: new contacts, and addresses already on the list (still ErrorCode 9, nothing changed). The public signup form, the re-subscribe action in the user area and subscribe requests sent by email pass UpdateIfUnsubscribed=true, so a person who subscribes again through them is re-subscribed as before. The journey Subscribe to list action and incoming subscription webhooks now skip a contact who unsubscribed from the target list.

Who needs to act: an integration that calls subscriber.subscribe and expects an unsubscribed contact to come back. Handle ErrorCode 12 as "skipped, the contact opted out". Send UpdateIfUnsubscribed=true only when the contact has given fresh consent, for example by submitting your own signup form.

subscriber.subscribe refuses a ListID set that spans accounts, and every list honours double opt-in ​

This is a security fix. subscriber.subscribe needs no authentication and accepts a comma-separated ListID. Before v6.0.6 only the first list was handled normally. Every other list in the set was written by ID alone: it could belong to any account on the install, and the address was always written as Subscribed with no confirmation email, even on a double opt-in list. Anyone could therefore add confirmed contacts to another account's lists through the public signup form or api.php (octeth/oempro#3956).

Two changes:

  • When any list after the first belongs to a different account than the first list, the call is refused with the new ErrorCode 13 and nothing is written to any list, the first one included. A list ID that does not exist is still skipped, as before.
  • Every list after the first now follows its own opt-in mode. A double opt-in list gets the address as Opt-In Pending and sends its own confirmation email. A single opt-in list gets it as Subscribed, as before.

Not affected: a single ListID, and a set of lists that all belong to one account on single opt-in. The success response and RedirectURL (taken from the first list) are unchanged.

Who needs to act: an integration or signup form that sends several lists of different accounts in one call. Send one call per account instead. A form that subscribes to several double opt-in lists of one account now sends one confirmation email per double opt-in list, and the contact is only Subscribed on each list after confirming it.

subscriber.subscribe refuses a ListID set of more than 10 lists ​

This is a security fix. Because every double opt-in list in a comma-separated ListID now sends its own confirmation email (see above), one unauthenticated call with a long set of lists could mail one address once per list. Before v6.0.6 nothing limited the length of the set, and every ID in it cost a list lookup, IDs that do not exist included (octeth/oempro#4006).

The set is now validated before any list is looked up:

  • Each ID is trimmed. Empty entries are dropped, and a repeated ID counts once (the first occurrence keeps its place, so the first list stays the primary list).
  • A set with more than 10 distinct lists is refused with the new ErrorCode 14. Nothing is written and no email is sent.
  • An ID that is not made of digits only is refused with ErrorCode 4 (invalid list). Before v6.0.6 such an ID after the first was skipped.

Not affected: a set of up to 10 lists, which behaves as described in the section above.

Who needs to act: a signup form or integration that sends more than 10 lists in one call. Split the lists across calls, or raise the limit with SUBSCRIBE_MAX_LISTS_PER_REQUEST in .oempro_env (see Octeth Configuration, "Public Subscribe List Cap"). A form that sends a non-numeric list ID must remove it.

Campaigns ​

campaign.update refuses to schedule a campaign whose email cannot be sent ​

Before v6.0.6, campaign.update scheduled or resumed a campaign whose email had no subject, no content or no From address, and returned Success: true. The send engine then failed every recipient and the campaign still ended as Sent (octeth/oempro#3945).

Such an update is now refused with the new ErrorCode 32, and the campaign is left unchanged. The ErrorText names each email and what it lacks, for example The campaign cannot be sent: email 41 has no subject, no HTML or plain content. The check runs when the update moves the campaign to a scheduled Ready, Sending or Pending Approval, or changes the email of a campaign that already is, and covers every A/B variation. An empty FromEmail is accepted when the delivery server (including a global routing segment's) enforces a From address or every target list has a sender email address. campaign.resume returns the same error.

Who needs to act: an integration that creates an email and schedules its campaign in one flow. Check the email.update response before calling campaign.update, and handle ErrorCode 32 by completing the email.

A campaign whose recipients all failed ends as Failed ​

Before v6.0.6, a campaign ended as Sent even when every recipient failed, and it was counted as a sent campaign for the account's payment period and statistics. Now a campaign that sent to no recipient and failed at least one (seed list emails are not counted) ends with CampaignStatus Failed and CampaignStatusReason 018, and is not counted as sent. When the failed run belongs to a repeating campaign, the repeating campaign is set to Failed with the same reason too, so it stops instead of failing again every cycle. A campaign with at least one sent recipient still ends as Sent, with its failures in TotalFailed.

Who needs to act: an integration that polls a campaign until it is Sent. Treat Failed as a final status too.

campaigns.delete no longer deletes a campaign that is sending ​

campaigns.delete skips every campaign whose status is Sending when the delete runs, and deletes the others in the list. The response is unchanged: it still answers Success: true, and the skipped campaign simply stays, with its statistics, recipients and queue (octeth/oempro#3791). Before, a sending campaign was deleted with its data while delivery went on, which could leave a partly delivered campaign with no report. The same rule applies in the classic interface and the new interface, which delete through the same command.

The change also closes a race: a scheduled campaign that became due in the same moment it was deleted could start sending after its rows were removed. Now exactly one of the two wins. Either the campaign is deleted and never starts, or it starts sending and the delete leaves it.

Auto-resend campaigns that campaigns.delete deletes together with the requested campaign are checked one by one, so an auto-resend campaign that has not started is still deleted when its original campaign is sending and kept.

Not affected: campaigns in any other status (Draft, Ready, Paused, Pending Approval, Sent, Failed). Deleting a user account (users.delete) still deletes all of the account's campaigns, sending ones included. AI agent keys were already refused a sending campaign (octeth/oempro#3781) and still are.

Who needs to act: an integration that deletes a campaign while it sends and expects it gone. Pause the campaign (campaign.pause) and then delete it, or wait until it has finished, and check with campaign.get that it no longer exists.

SMS campaigns ​

Before v6.0.6 every tracked link in an SMS campaign carried a 20-character code, both in /sc/ short links and where {{ SMS:TrackingCode }} is replaced in a passthrough link. Twenty characters often pushed a message into a second billed part. New messages now carry a five-character code of letters and digits, for example https://go.example.com/sc/aB3x9 (octeth/oempro#3935). The cost estimate counts five characters per link, so it matches what is sent.

A five-character code is case-sensitive and resolves for 90 days after the message is sent, or for the campaign's link expiry (LinkExpiryHours) when that is longer. After that its click and its Keitaro conversions are no longer attributed. Codes already sent keep their 20 characters and keep working. A test message carries the code 00000, which never resolves.

Not affected: the Keitaro postback URL, the poller and the postback parameters. Messages already queued before the upgrade are sent with five-character codes, so they can use fewer parts than were estimated, never more.

Who needs to act: an account that forwards {{ SMS:TrackingCode }} links through a Cloudflare redirect rule with len(http.request.uri.path) eq 21. Change the rule before you send a campaign after the upgrade, as described in shorter links with Cloudflare, or new links stop reaching Keitaro. Anything else that validates the code's length or pattern must accept five characters as well as 20.

Email Gateway ​

  • EmailGateway.SendEmail and the SMTP relay now refuse a send whose From address is on the administrator's forbidden From addresses list (staff Settings > Preferences, Lists and senders section). EmailGateway.SendEmail answers HTTP 422 with {"Errors": [{"Code": 45, "Message": "From email address is forbidden by the administrator"}]}. The SMTP relay answers 550 5.7.1 From address is forbidden by the administrator (emailgateway.smtprelay returns ErrorCode [14]). Nothing is queued and no credit is charged, and each refusal is recorded as a rejected-by-oempro event with the reason "The From address is forbidden by the administrator." and logged at warning level. The address is checked as the caller wrote it and as it is sent on the sender domain, and for the SMTP relay both the header From and the envelope MAIL FROM are checked. The rules are the ones the email editor uses: matching ignores case, a line without * must equal the whole address, and * matches any text. Before, the list did not apply to the email gateway, so an account refused an address in the email editor could still send as it through the gateway. Journey Send email steps and AI agent key sends go through EmailGateway.SendEmail and follow the same rule. Installs with an empty list see no change. There is no opt-in flag, because the old behavior let gateway senders bypass the operator's list (octeth/oempro#3623).
  • The SMTP relay now delivers a message to its SMTP envelope recipients (the RCPT TO commands) and to nobody else, as a mail server does, and keeps the To and Cc headers exactly as the sender wrote them. Before, it delivered to the first RCPT TO address plus every address in the Cc header, and to header Bcc addresses: a header Cc address the client left out of the envelope still got the message, and an envelope address that also appeared in the To header (other than the first) was never delivered. A split-delivery setup, where your own mail server delivers internal recipients itself and relays only the external ones through Octeth, therefore sent internal recipients a second copy. Envelope recipients that the headers do not name are delivered as blind copies and never appear in a header. Messages sent from Gmail, which were already delivered to the envelope, no longer list blind-copy recipients in the visible To header. Credits are charged per envelope recipient. With the PowerMTA send method the pickup file now carries one x-receiver line per envelope recipient (before, only the first RCPT TO address was written, so the other recipients were never delivered), and with the SaveToDisk send method a relayed message's file now starts with the same x-sender and x-receiver lines. On an install whose delivery uses the local MTA or PHP mail() send method, headers are rebuilt from the delivered recipients, so header addresses outside the envelope drop out of the visible To and Cc. EmailGateway.SendEmail (the HTTP API) and journey sends are not affected. There is no opt-in flag: a standard SMTP client sends every To, Cc and Bcc address in the envelope, so it sees no change (octeth/oempro#1314).
  • A hard bounce of an email gateway message now suppresses the recipient for every account on the install, not only for the account that sent it. Before, the gateway bounce handler wrote an account-wide suppression entry (the sending account's UserID, list 0), so other accounts kept sending to an address the receiving server had already rejected as permanently undeliverable. It now writes one system-wide entry (RelOwnerUserID = 0, RelListID = 0, source Hard Bounced), the same entry campaign bounces processed by PowerMTA already wrote. EmailGateway.SendEmail and the SMTP relay refuse a suppressed recipient as before, now for every account. Two consequences for the sending account: an address on its whitelisted email addresses is no longer exempt from a gateway hard bounce, and it cannot remove the entry from its own suppression list, because only an administrator can delete a system-wide entry. Soft bounces are still ignored, and entries written before the upgrade are not changed. There is no opt-in flag: a hard bounce is a property of the mailbox, and the account-wide entry let every other account keep hitting a dead address (octeth/oempro#3946).

Tier 2: shape and value changes ​

Subscribers ​

subscriber.subscribe returns the status code instead of ErrorCode 7 for a contact's own unique value ​

Before v6.0.6 the uniqueness check of a unique custom field counted the contact's own stored value, so a call for an address already on the list that carried its own unique value failed with ErrorCode 7, and so did an address joining a second list with a global unique value it already held (octeth/oempro#3883). The check now ignores the contact's own record. Such a call returns ErrorCode 9 (already on the list) or ErrorCode 12 (unsubscribed), or succeeds for a new list or with UpdateIfUnsubscribed=true. A value held by a different subscriber still returns ErrorCode 7.

Who needs to act: an integration that treats ErrorCode 7 from subscriber.subscribe as "this contact already exists". Handle ErrorCode 9 and ErrorCode 12 for that case.

subscribers.get returns only a segment's members for a segment builder segment ​

Before v6.0.6, subscribers.get with a numeric SubscriberSegment returned a page of the whole active list for a segment built with the segment builder, segment.create with RulesJson or the MCP create_segment tool, while TotalSubscribers counted only the segment. Legacy campaign recipient estimates and segment exports had the same fault (octeth/oempro#3943). These reads now return only the segment's members. With SearchKeyword and a numeric SubscriberSegment, the call now searches within the segment and TotalSubscribers counts the matches, where before it returned no subscribers. A segment of another account, or one whose RulesJson is malformed, reads as empty. When the segment cannot be resolved, the call returns the standard error envelope instead of a page of the wrong subscribers, and a segment export is marked failed.

Not affected: segments with classic SegmentRules, and calls without SubscriberSegment.

Who needs to act: an integration that pages a segment through subscribers.get. Expect fewer rows, matching TotalSubscribers, and handle the error envelope.

User groups ​

usergroup.update and usergroup.patch refuse an archived or missing delivery server ​

Before v6.0.6 both commands stored any TargetDeliveryServerID_* value in Options and answered success, even for a delivery server that does not exist or is archived. Sending then fell back to the system default without telling the admin (octeth/oempro#3847). Such a value is now refused with ErrorCode 37 (usergroup.update) or ErrorCode 41 (usergroup.patch), with ErrorField naming the key, and nothing is written. This is the rule usergroup.options.patch already applied as ErrorCode 7.

Not affected: 0 (system default), an active delivery server, and calls that do not send Options.

Who needs to act: an integration that sets a group's delivery servers through these commands. Handle ErrorCode 37 or 41, and pick an active server or 0.

SMS reporting ​

smscampaign.stats.breakdown counts each message once for carrier and country ​

Before v6.0.6 the carrier and country dimensions grouped every event of the campaign by the raw column. A message's carrier and country come from its delivery report, so every message was counted under the empty value through its queued, send_attempt and sent events, and each message whose report named a network was counted a second time under that network. A 500-message campaign could report 501 messages with no carrier plus 22 under named networks (octeth/oempro#3920). Each message is now resolved to one value first, the one its latest delivery report (delivered, undelivered or expired) named, so the rows add up to the campaign's messages and are ordered by Messages. A click's geolocated country no longer counts toward the country breakdown. Events for these rows is the number of events of the messages in the row.

Every response also carries the new TotalMessages field, the message count independent of Limit.

Not affected: the hour, day, gateway_status, error_code, link and sender_id dimensions, which still count events per value.

Who needs to act: an integration that reads carrier or country breakdowns. The empty row now holds only messages with no reported value, and the figures under each network no longer include a second count of the same message.

Keitaro conversions count status sale only ​

Before v6.0.6 any latest status except rejected was a conversion, so lead and custom statuses such as quiz_started counted in smscampaign.stats Integrations.Conversions, Integrations.Revenue and Integrations.Links, in the integration_conversion series of smscampaign.stats.timeseries, and in the segment integration converted and not integration converted operators without a status. Now only sale counts (octeth/oempro#3925). The rule is applied when reading, so existing campaigns show fewer conversions and less revenue after the upgrade, with no data change.

Not affected: every status is still stored, listed in Integrations.Statuses and returned by smscampaign.events.browse, and a segment rule that names a status still matches that status.

Who needs to act: an integration that reads Keitaro conversions or revenue. Expect lower figures where Keitaro sends lead or custom statuses, and name status in a segment rule to target a status other than sale.

SMS campaigns ​

smscampaign.get says whose quiet hours a campaign follows ​

The QuietHours block carries the new Source field: campaign when the campaign sets both QuietHoursStart and QuietHoursEnd, default when it follows the install's quiet hours (octeth/oempro#3919). Before v6.0.6 the block gave only the window in force, so a client showed the install default as if the campaign had chosen it.

Not affected: every other field of smscampaign.get, and when messages are held. The sender resolves quiet hours exactly as before.

Who needs to act: nobody. A client that labels a campaign's quiet hours can read Source to mark the install default.

Campaign test sends: two install switches, and two new sms.testnumbers.get fields ​

sms.testnumbers.get returns two new Limits fields, VerificationRequired and LimitsEnabled. Both are true unless an operator switches them off with SMS_TEST_REQUIRE_VERIFIED_NUMBER or SMS_TEST_LIMITS_ENABLED (or the staff SMS settings). With both on, smscampaign.test behaves exactly as in v6.0.5. With verification off it accepts any valid phone number and no longer returns error 14, and with limits off it no longer returns errors 16 and 17. admin.sms.settings.get lists 19 settings instead of 17.

Not affected: installs that leave both settings at their default.

Who needs to act: nobody. A client that explains test-send refusals can read the two new fields.

Journeys ​

Journey emails skip the email API request limit and wait out throttles and account holds ​

Before v6.0.6 a journey Send email spent the account's POST /api/v1/email request budget (100 requests a minute), shared with the account's own transactional API traffic, so a large enrolment failed most of its sends with HTTP 429 and could also refuse the account's own API sends (octeth/oempro#3944). A journey send signed by Octeth now skips that limiter. It is still bounded by the account's email gateway send rate limit, credits and daily cap, and unsigned requests are limited as before.

A journey Send email refused with HTTP 429, or refused because of the account itself (codes 12, 32, 34 and 40: disabled account, inactive sender domain, unverified phone, pending approval), used to count as a failed attempt, so after six attempts subscribers who passed the step while the account was held never got the email (octeth/oempro#3951). These refusals now wait without spending an attempt: a throttle retries after the gateway's Retry-After (60 seconds when none is given) for up to 24 hours, and an account hold retries hourly for up to 7 days, then the entry is dead-ended with a reason naming the throttle or hold. A sender domain that is not Enabled when the step runs (Approval Pending, Suspended, Blocked or Disabled) used to skip that email and move the subscriber on. It is now the same domain hold and waits. A retry is never scheduled past a wait cap. The recorded ErrorCode on the action execution is now Throttled or AccountHold:<reason> instead of SendFailed, and retries after an ordinary failure carry up to 20 percent jitter.

journey.get takes the new opt-in IncludeSendHolds parameter, which adds a SendHolds array listing the holds this journey's entries are waiting on.

Not affected: entries already dead-ended before the upgrade, which are not revived, and journey.get calls without IncludeSendHolds.

Who needs to act: an integration that reads journey action ErrorCode values. Expect Throttled and AccountHold:<reason> for these refusals.

Segments ​

Changing a segment's rules resets its count timestamp ​

Before v6.0.6, segment.update and the legacy segment editor kept the stored count when a segment's rules changed, so segments.get served the old rules' count as fresh for up to 10 minutes (octeth/oempro#3939). When the rules or the match operator change, SubscriberCountLastCalculatedOn now becomes 0000-00-00 00:00:00 and the cached count is dropped. SubscriberCount keeps the last number until the recount. The legacy editor used to write a timestamp three days in the past instead. A count is now saved only while the segment still has the rules it was counted for, so a count that runs across a rule edit is dropped and the segment is counted again, and opening a segment in the legacy editor no longer stores 0 when its recount times out. Subscriber deletions and imports still mark every counted segment for a recount, but leave a never-counted segment never counted.

segments.get also takes the new opt-in ComputeMissingCount parameter. With it and one SegmentID, a never-counted segment is counted during the request (up to SEGMENT_COUNT_INLINE_TIMEOUT, default 10 seconds). The AI connector's get_segment_count, create_segment and list_segments tools report CountStatus (counted or pending), and a segment still being counted has SubscriberCount: null, never 0 (octeth/oempro#3937).

Not affected: a rename or an update that leaves the rules as they are, and segments.get calls without ComputeMissingCount.

Who needs to act: an integration that reads SubscriberCount right after changing a segment's rules. Treat a zero SubscriberCountLastCalculatedOn as "not counted yet", or pass ComputeMissingCount=true with the SegmentID.

Bounce and unsubscribe attribution ​

Before v6.0.6 bounces processed from the bounce spool, cli/bounce.php and the internal bounce register call were stored with no email (A/B variant) and did not count toward the sender domain's bounce total, and unsubscribes through the unsubscribe link were stored with an empty channel (octeth/oempro#3947). Such a bounce now records the email it was sent from when that email belongs to the account, and counts toward its sender domain. Such an unsubscribe records List-Unsubscribe one-click for an RFC 8058 one-click request from a mailbox provider, or Unsubscribe link for a link confirmed on the unsubscribe page, next to the existing List-Unsubscribe header and SPAM complaint values in the subscriber's activity history.

Not affected: rows recorded before the upgrade, and unsubscribes posted directly to a customer-hosted form, which keep an empty channel.

Who needs to act: nobody. A report that reads the unsubscribe channel or per-email bounces will see the new values on new rows.

Hard-bounce suppression entries ​

A hard bounce writes one system-wide suppression entry ​

Before v6.0.6 a hard bounce processed through the internal bounce register call, cli/bounce.php, the bounce spool, cli/pop3_bounce.php, the Amazon SES plugin or manual bounce processing wrote two suppression entries for the same address: one account-wide (RelOwnerUserID = the sending account, RelListID = 0) and one system-wide (RelOwnerUserID = 0, RelListID = 0). A soft bounce promoted to hard by the soft bounce threshold wrote only the account-wide entry. Every hard bounce, attributed to a campaign or not, and every threshold promotion now writes exactly one system-wide entry with source Hard Bounced (octeth/oempro#3946).

Not affected: whether an address is suppressed. Both old entries already suppressed it for the sending account, and the system-wide entry already suppressed it for every other account. A promoted soft bounce, which was suppressed for the sending account only, is now suppressed for every account. Entries written before the upgrade are not removed, so an existing install keeps its duplicate pairs.

Who needs to act: an integration that counts raw rows of the suppression list, or reads suppression.browse expecting a hard-bounced address on the account's own scope. Expect the address once, as a system-wide entry. An address on an account's whitelisted email addresses was never exempt from a campaign hard bounce, because the system-wide entry was written anyway, and this is now the documented rule.

Account suspension by the abuse protection plug-ins ​

Peak Monitor, SpamAssassin Content Validator and PhishGuard now suspend an account by setting its AccountStatus to Disabled, so user.get and the user lists show the account as disabled and it can no longer sign in or send through the email gateway. Before, they only called a WooCommerce endpoint, and on an install without the WooCommerce plug-in a suspension changed nothing. New installations enable the three plug-ins with non-empty defaults. Upgraded installations keep their enabled plug-ins and saved settings until an administrator applies the new defaults. See Abuse Protection (octeth/oempro#3085).

Not affected: API request and response fields. The new suspension source, reason and time columns are not returned by any endpoint. Enabling the account again with user.update (AccountStatus=Enabled) clears them.

Who needs to act: an integration that disables and re-enables accounts itself and treats every Disabled account as one it disabled. A plug-in can now disable an account too.

Tracking flags ​

OpenTracking and LinkTracking accept more values on email.update and the sender-domain endpoints ​

email.update, user.senderdomain.create, user.senderdomain.update and emailgateway.updatedomain now read OpenTracking and LinkTracking with one shared set of values (octeth/oempro#3950). On: true, 1, "true", "1", "yes", "on", "enabled". Off: false, 0, "false", "0", "no", "off", "disabled", "". Strings are case-insensitive and surrounding whitespace is ignored. Omitting the parameter or sending null keeps each endpoint's default, unchanged.

What changes for existing callers:

  • email.update: "yes", "on" and "enabled" (any case) turned tracking off before, and now turn it on.
  • user.senderdomain.create and user.senderdomain.update: "enabled" turned tracking off before, and now turns it on. "yes" and "on" already meant on.
  • emailgateway.updatedomain: "true", "yes", "on" and "enabled" turned tracking off before, and now turn it on.

A value outside both lists keeps the result it had before and writes a warning naming the field to the application log. Send StrictFlags=true to have such a value refused instead, with a per-endpoint error code listed in the API reference. Without StrictFlags no request that succeeded before fails now. Other on/off parameters on these endpoints, such as UnsubscribeLink and UseTrackingSubdomain, are not changed.

Who needs to act: an integration that sends one of these words expecting tracking off. Send false or 0 instead.

Upgrade checklist ​

  • [ ] If you call subscriber.subscribe, handle ErrorCode 12 (the contact unsubscribed from the list, nothing changed). Send UpdateIfUnsubscribed=true only for a contact who asked to subscribe again.
  • [ ] If you call subscriber.subscribe with more than 10 lists in ListID, split the call or raise SUBSCRIBE_MAX_LISTS_PER_REQUEST. A larger set is refused with ErrorCode 14, and a non-numeric list ID with ErrorCode 4.
  • [ ] If you call subscriber.subscribe with unique custom fields, ErrorCode 7 now means another subscriber holds the value. An existing contact resubmitting its own value gets ErrorCode 9 or ErrorCode 12.
  • [ ] If you call campaigns.delete for campaigns that may be sending, pause them with campaign.pause first or wait until they finish. A Sending campaign is skipped and the call still answers Success.
  • [ ] If you set TargetDeliveryServerID_* through usergroup.update or usergroup.patch, handle ErrorCode 37 or ErrorCode 41 (the server is archived or does not exist, nothing changed).
  • [ ] If you read smscampaign.stats.breakdown by carrier or country, expect each message in one row (the rows add up to TotalMessages) and a smaller empty Dimension row.
  • [ ] If you read Keitaro Conversions or Revenue, expect sales only, and set status in a segment rule to match lead or a custom status.
  • [ ] If you forward {{ SMS:TrackingCode }} links through a Cloudflare redirect rule, change its length check to accept 6 as well as 21 before you send a campaign. Anything else that checks the tracking code must accept five characters.
  • [ ] If your forbidden From addresses list (staff Settings > Preferences, Lists and senders section) holds wildcard lines, check that none of them matches a From address your accounts send transactional email from through the email gateway. Such sends are refused after the upgrade (Code 45 on EmailGateway.SendEmail, 550 5.7.1 on the SMTP relay).
  • [ ] If an SMTP client relays through Octeth with an envelope that leaves out some header To or Cc addresses, check that this is intended. Those addresses no longer receive the message from Octeth after the upgrade.
  • [ ] If an account relies on its whitelisted email addresses or on removing addresses from its own suppression list to keep mailing an address that hard-bounced through the email gateway, note that a new gateway hard bounce now writes a system-wide entry that only an administrator can remove.

Previous releases: v6.0.5, v6.0.2, v6.0.1, v6.0.0

Any questions? Contact us.